Download Agreement - General Guide To Personal and Societies Web Space

Agreement: A
Relational Property or a
Functional Projection?
Evidence from Romance
A thesis submitted by
Sandra Paoli
to the University of York
in fulfilment of the degree of
Masters by Research in Linguistics
September 1997
1
Introduction
Pollock’s (1989) analysis of the difference between English and French with respect to
adverbial placement has led to the formulation of a functional projection, AgrS, to which
French, but not English, verbs move.
Kayne (1989) investigating object - past participle agreement in French proposes an
analysis that sees the agreement contingent upon a government relation between a
functional projection Agr and the DP object, thus drawing a parallel between subjectverb and verb-object agreement.
Chomsky (1991), speculating on Kayne’s work, suggests the existence of two functional
nodes responsible for the instantiation of agreement, AgrS, for subject agreement, and
AgrO, for object agreement.
More recently, Chomsky (1995) has proposed to eliminate agreement functional
projections altogether, so that only those functional projections with clear semantic
content - such as Tense or Negation - were allowed in the system, but has not formulated
how to capture the empirical effects.
The tension between the notion of agreement as a pure relation and agreement as a
functional projection has been mentioned increasingly frequently in footnotes but the
question has rarely been faced directly.
The aim of this dissertation is two-fold. First, to provide a unified structural account of
past participle agreement (PtPPL) with a full post-verbal DP (pvDP) and adjectival
agreement. Secondly, on a more theoretical level, to make a small contribution to the
understanding of the notion of agreement.
An interesting area in which Romance languages compare interestingly to each other
and to English is the formation of the active form of compound tenses. In English tenses
such as the present or past perfect are constructed by a form of the auxiliary to have,
present or past respectively, followed by the PtPPL of the main verb.
In contrast, in the Romance languages the selection of the auxiliary ranges over two
verbs, to have (A aux) and to be (E aux). Spanish and Rumanian, like English,
make use exclusively of the A aux in the active form, while the E aux is used only in
2
the passive constructions. Various Central-Southern Italian dialects and some
varieties of Catalan, on the other hand, display only the use of the E aux for both
active and passive constructions. Standard Italian (SI) and French find themselves
in a middle position and make use of both auxiliaries in active constructions.
Many attempts have been made to account for this alternation in French and SI
(Leone (1954), Leone (1970), Agard (1977), Perlmutter (1980), Hyams (1981),
Burzio (1981, 1986), Lefebvre (1988), Taraldsen (1987), Lois (1990), Vikner &
Sprouse (1988), Guéron & Hoekstra (1988), Vikner (1991), Kayne (1993), Cocchi
(1994, 1994a, 1994b)). Leone (1970: 30) clearly states that the choice of auxiliary is
dependent on the nature of the PtPPL, in particular, on whether the speaker’s
“linguistic conscience” felt the PtPPL to act as an adjective or not. This dissertation
will not deal with auxiliary selection phenomena; nevertheless, the topic will be
touched upon when relevant to the main topic, PtPPL agr.
In French the E aux is used with reflexive verbs and verbs used reflexively - il s’est
ennuyé “he got bored”, il s’est lavé “he washed himself” - and some movement
verbs il est arrivé “he has arrived”, il est allé “he has gone”. Impersonal and
transitive verbs use the A aux and so does être (to be).
In SI the E aux - essere - is used with reflexive verbs and with verbs used
reflexively, with si-impersonal constructions and with all those movement verbs
which fall in the so-called category of unaccusative verbs, i.e. those verbs whose
structural subject is generated post-verbally, in the canonical object position
(Perlmutter (1978), Burzio (1981, 1986)). The A aux - avere - is usually used with
transitive verbs. There are some verbs which admit both auxiliaries, A and E with
the choice reflecting a clear semantic difference , and some intransitive verbs that
select the A aux. Some examples are given in (1). (1) a is a si-impersonal
construction, (1) b and c show some verbs which allow for the alternation between
the two auxiliaries, and (1) d an intransitive verb selecting the A aux:
(1) a. Negli ultimi
tempi
si é/*ha
molto
sentito
In-the
last
times
SI
is/*has a-lot
heard
parlare di esperimenti di ingegneria
genetica
speaking of experiments of Engineering Genetics
“Recently, there was a lot heard about Genetic Engineering experiments”.
3
b.
Sono corsa/*Ho
corso
a casa
I-am run /*have run
at home
“I have run home (i.e. towards home)”
c.
Ho corso /*Sono corsa per un’ ora
Have run /*am run
for an hour
“I have run (i.e. jogged) for an hour”
d.
Il cane ha/*é
starnutito
The dog has/*is sneezed
“The dog has sneezed”
Traditional grammars such as Lepschy & Lepschy (1981), provide the reader with a
straightforward rule: avere (A) aux is to be used with all transitive verbs and essere
(E) aux with all intransitive verbs. Intransitives selecting avere are considered
exceptions and are grouped in a list: camminare (to walk), chiacchierare (to chat),
piangere (to cry), respirare (to breathe), ridere (to laugh), sorridere (to smile),
tossire (to cough), and so on.
An issue closely connected with auxiliary selection and which is often discussed
together with it is the agreement instantiated between a DP base generated postverbally (pvDP) and the PtPPL (Leone (1970), Burzio (1986), Lefebvre (1988),
Taraldsen (1987), Lois (1990)). Burzio (1986) formally expresses the link between
E aux selection and PtPPL agr by accounting for them in a unified manner (see
chapter 1 section 1.3 for a detailed discussion of his analysis).
In SI PtPPL is instantiated whenever the E aux is selected. A aux selection and
PtPPL agr with the PtPPL’s internal argument is triggered only when the DO
undergoes cliticisation and moves to a higher position, as shown in examples (2) a,
b and c; if not movement takes place, the unmarked masculine, singular (m, sg)
form ending in “-(x)to” surfaces - this will be always referred to as (UM) for
unmarked:
(2) a.
Paola
ha mangiato
Paola
has eaten (UM)
“Paola has eaten the apple”
la mela
the apple (f, sg)
4
b.
?*Paola ha mangiata
Paola
has eaten (f, sg)
“Paola has eaten the apple” 1
la mela
the apple (f, sg)
c.
Paola
l’
ha mangiata
Paola
it has eaten (f, sg)
“Paola has eaten it”
t
t (f, sg)
i
i
i
i
Because of the co-occurrence of movement and PtPPL many analyses (Kayne
(1985, 1989), Burzio (1986), Taraldsen (1987), Lois (1990)) have taken the two
phenomena to be interdependent.
When we turn to Italian dialects, the behaviour with respect to both auxiliary
selection and PtPPL agr is extremely varied.
Friulian, the dialect on which this work is based, is spoken by over 700,000 people
(Cf Vanelli (1997)) in the region of Friuli Venezia Giulia, Italy. The variety
investigated here is a central-southern variety of Friulian, spoken south of Udine,
in the areas between Palmanova and Aquileia, in particular in the village of
Castions delle Mura. Together with the Ladin dialects of the Dolomitic Alps in
South Tyrol - Italy - and the Romansch dialects of Grisons - Switzerland – Friulian
The judgement here is not strictly ungrammatical, but it is felt to be archaic
by most speakers. Nowadays there are still people using the agreeing form of the
PtPPL in sentences like (2) b but the phenomenon is restricted mainly to highly
educated, upper class speakers, especially those who are or have spent a long time
in academic institutions. This suggests that this is learnt rather than spontaneous.
However, since this is not widespread enough to be relevantly significant - to the
best of my knowledge there are no very young speakers, i.e. under the age of 20,
would spontaneously produce it - it will here be left aside. It will be maintained
that the non agreeing form of the PtPPL in contexts not having cliticisation will be
the natural choice for all speakers.
1
5
is one the so-called Rhaeto-Romance languages. It appears in many different
varieties depending on the areas in which it is spoken, and it is influenced by the
nature of the surrounding languages. Examples of this influence can be found in
the varieties spoken in the area adjacent to the Carnic Alps along the Austrian
border and in the areas around Gradisca near the border with Slovenia which
displays features of the language across the border.
As with all other dialects, Friulian compares very interestingly to SI and in a way it
can be considered more conservative than the latter. Here, too, the choice of
auxiliary ranges over the E aux and the A aux, but the types of verbs selecting the
one or the other is partly different form SI (for a detailed account the reader is
referred to Benincà & Vanelli (1984)) - and there is variation between the different
varieties of Friulian.
A challenge to the conditions on the agreement instantiated in SI by a transitive
verb selecting the A aux and its direct object (DO) comes from Friulian. Friulian is,
in some respects, more traditional than SI, allowing for constructions which were
allowed in XIX century Italian and but have now disappeared from SI and are
generally considered archaic. The most salient of these constructions is given in
(3): with transitive verbs selecting the A aux, agreement is triggered between the
PtPPL and a DO in situ, i.e. which has not undergone movement - or at least not
visibly:
(3) a.
b.
Marie
e
a
mangiadis li caramelis
Mary SCL has eaten (f, pl)
the sweets (f, pl)
“Mary has eaten the sweets”
Toni el a mangiâs
Tony SCL has eaten (m, pl)
“Tony has eaten the pears”
i
pirus
the pears (m, pl)
Given that from the linear order the DO does not seem to have moved from its
base-generated position, it is difficult to account for the Friulian data in terms of a
theory invoking movement.
6
This dissertation is structured as follows.
In chapter 1 I will present and discuss some ideas as to why PtPPLs seem to
behave similarly to APs. Chapter 2 presents an in-depth evaluation of different
semantic accounts of PtPPL agr and shows that even with extensive modification a
semantic analysis of the phenomenon cannot deal adequately with Friulian
transitive activity verbs with an indefinite object. Chapter 3, presents a new
analysis, where empirical evidence supports the complete elimination of an
agreement functional projection responsible for PtPPL agr - be it AgrO or AgrP -.
The semantic affinity between PtPPLs and APs is then interpreted on a syntactic
level by adopting the same mechanical process for both PtPPL and DP internal agr,
namely feature copying. The differences between the two categories are captured,
syntactically, by the different operational domain of the feature-copying process.
The triggering factor that underlies the two types of agreement is the matching of
[+N] features. Friulian PtPPL is assumed to be specified for [+V,+N] features,
while SI PtPPL is claimed to carry [+V] features. This claim finds support from a
diachronic investigation of the evolution of the PtPPL and the A aux from Latin to
the modern Romance Languages presented in chapter 4. Here the differences
noticed between younger and older Friulian speakers are explained claiming that
Friulian is still undergoing the developmental process already accomplished by SI.
7
Chapter 1
Past Participle Agreement - Syntactic
Accounts
1.0
1.1
1.2
1.3
1.4
1.5
1.5.1
1.6
Introduction
Past Participle Agreement: the facts in Friulian
What is Friulian?
Burzio (1986) and its adaptation in Paoli (1996)
Problems with Paoli (1996)
Lehmann’s (1988) internal vs external agreement
[gender, number] features: Giorgi and Pianesi (1991)
Conclusion
1.0 -
Introduction
This chapter will concentrate on the data and present some questions left
unanswered by two previous analyses, Burzio (1986) and Paoli (1996). In section
1.1 it is shown that there are no differences between SI and Friulian with respect to
PtPPL agr when the E aux is selected, so that this work will concentrate only on the
A aux. Section 1.2 presents some background information on Friulian and refers to
Paoli (1996), where the appearance of subject clitics (SCL) and PtPPL agr with a
pvDP in situ were accounted for by formulating the “Agreement Parameter”.
Section 1.3 digresses into a detailed analysis of Burzio’s (1986) and the adaptation
of this made in Paoli (1996). Some problems left unanswered by the two analyses
are illustrated in section 1.4. The feature specification of PtPPL and adjectival
agreement suggests an affinity between PtPPLs and APs, captured in section 1.5 by
8
Lehmann (1988) and in Section 1.5 by Giorgi and Pianesi (1991).
1.1 -
Past Participle Agreement: the facts in Friulian
Both in Friulian and SI, as can be seen respectively from the examples in (4) and
(5), the E aux triggers PtPPL agr:
(4) a.
Maria
e
Lucia
e
Mary
and Lucy SCL 2
“Mary and Lucy have gone”
son ladis /
*lât
are gone(f, pl) /*(UM)
b.
Piero
e
Toni e
son rivâs /
*rivât
Piero
and Toni SCL are arrived (m, pl) /*(UM)
“Piero and Tony have arrived”
c.
Maria
e
Lucia
si
Mary
and Lucy themselves
“Mary and Lucy got washed”
b.
Piero
e
Toni
si
son scris /
*scrit
Piero
and Tony
to each other are written (m, pl)/*(UM)
“Piero and Tony have written to each other”
(5) a.
b.
son lavadis /
*lavât
are washed (f, sg) /*(UM)
Maria
e
Lucia
sono andate /
*andato
Maria
and Lucia
are gone (f, pl) /*(UM)
“Maria e Lucia have gone”
Piero
e
Toni
sono arrivati /
*arrivato
Piero
and Toni
are arrived (m, pl) /*(UM)
“Piero and Tony have arrived”
SCL stands for Subject CLitic. See section 1.3 for an account of their nature
and function.
2
9
c.
Maria
e
Lucia
si
sono lavate / *lavato
Mary and Lucy themselves are washed (f, pl) /*(UM)
“Mary and Lucy got washed”
d.
Piero
e
Toni si
sono scritti /
*scritto
Piero
and Toni to each other are written (m, pl) /*(UM)
“Piero and Tony have written to each other”
This work will concentrate only on transitive verbs with a DO in situ selecting the A
aux.
Examples (2) a, b and c show that in SI PtPPL agr with a DO in situ is not usually
triggered. This is, on the other hand, what happens in Friulian, as shown in
example (6) a, where PtPPL agr is triggered with a DO in situ as well as, on a
parallel with SI, when the DO becomes a clitic, as shown in (6) b and d:
(6) a.
Piero
el a
serâs /
serât i
barcôns
Piero
SCL has shut (m, pl) / (UM) the windows (m, pl)
“Piero has shut the windows”
b.
Piero
ju
a
Piero
them
have
“Piero has shut them”
c.
Maria
e
a viodude / viodût to
Mary
SCL has seen (f, sg) / (UM) your
“Mary has seen your sister”
d.
Maria
le a viodude / *viodût t
Mary
her has seen (f, sg) /*(UM)
“Mary has seen her”
i
serâs /
*serât t
shut (m, pl) /*(UM)
i
i
sûr
sister (f, sg)
i
The possibility of having the UM form of the PtPPL in Friulian is witnessed in
younger speakers, while the older ones prefer the agreeing one. This could be due
to the fact that younger speakers are more in contact with SI speakers than the
10
older ones, having a wider network, or perhaps to a change in progress which is
showing up in the younger speakers or, still, to an interaction of the two. This issue
will be further developed in chapter 4. Throughout this work we will focus our
attention on the variety of Friulian spoken by older people. The one spoken by
younger people will be referred to when speculating on the possibility of a change
in progress.
Section 1.2 illustrates some general characteristics of Friulian.
1.2 -
What is Friulian?
Friulian, like other Northern Italian Dialects, has a set of subject clitics (SCL)
which co-occur with lexical and pronominal subjects. Examples (7) from Friulian
and (8) from SI show a comparison between the two languages:
(7) a.
Marie
e
Mary
SCL
“Mary sings”
ciante
sings
b.
Je
e
SheSP 3 SCL
“She sings”
ciante
sings
c.
pro
ciante
sings
d.
* pro
E
∅
SCL
“(She) Sings”
∅
(8) a.
b.
Ciante
Lei
SheSP
c.
Sings
Maria
Mary
canta
sings
canta
sings
pro
Canta
∅
Sings
SP stands for “strong pronoun”,i.e. those pronouns which can occur by
themselves in subject position as “she” She left.
3
11
Paoli (1996) investigated the nature of these subject clitics. The ungrammaticality
of (7) d compared to its Italian grammatical counterpart (8) c, led me to think that
Friulian was not one of the so-called pro-drop languages. However further
investigations showed that Friulian displayed syntactic properties shared by prodrop languages like Italian and thought of as being connected to the “pro-drop”
phenomenon (Rizzi (1982)): free subject-verb inversion and lack of “that-trace”
effects. The question to be answered was “what are these subject clitics?”, and an
analysis was proposed following Rizzi (1986). Friulian SCLs, as other SCLs in
Northern Italian Dialects, are different from French subject clitics (Cf Kayne 1975)
in that they are clitics by virtue of being heads, but are an expression of agreement,
and as such, base-generated under Infl. Consequently, I claimed that Friulian
agreement features are not only expressed as a morphological marker on the verb
that shows inflection, but are also phonetically realised as a SCL that agrees in _ features with the verb and its lexical or pronominal subject 4.
In the general approach to inflectional morphology developed by Baker (1985,
1988), morphologically complex words are derived from more basic elements roots, suffixes, stems - by means of a syntactic process of incorporation, a variant of
Move α affecting heads.
Following this approach the SCL was assumed to be base generated under Agr - in
a Pollockian (1989) framework - and the finite verb form adjoined to it through a
process of head-to-head movement obtaining the following structure for (7) a:
(9)
AgrS
Spec
AgrS’
Marie
e
j
Agr
TP
ciante
i
Spec
T’
t
VP
j
T
t
i
Spec
t
V
j
t
4
V’
i
Cf Poletto (1992), for a different analysis.
12
To express the difference between Italian which does not display these SCLs and
Friulian, it was claimed that in Friulian the AgrS node is specified for strong
features which need to be spelt out taking the form of SCLs.
Some more support for this was derived from the investigation of the instantiation
of agreement between a PtPPL selecting the A aux and its DO in situ.
In order to capture the link between the two manifestations of “extra” agreement,
Burzio’s account was considered alongside others (Kayne (1985), Bouchard (1985),
Lefebvre (1988), Brown (1988), Lois (1990), Cortés (1993), Taraldsen (1987)), and
amended. We need to digress slightly at this point to present Burzio’s analysis and
its adaptation in Paoli. Only the part regarding PtPPL agr will be considered in
some depth.
1.3 -
Burzio (1986) and its adaptation in Paoli (1996)
Fundamental to Burzio’s analysis is the distinction - put forward by Perlmutter’s
(1978) work formulated in a relational grammar framework - within the group of
intransitive verbs between unergatives and unaccusatives. Using as a test for
“unaccusativity” the fronting of the clitic pronoun ne (of it/them) Perlmutter
shows that the subject of verbs like telefonare - to telephone - is generated in a
different position from the subjects of verbs like arrivare - to arrive. More
specifically, while the former is base generated in the canonical position of the
subject, i.e. pre-verbally, the latter is base generated post verbally, and behaves like
a kind of object. This can be seen in the following examples: in (10) a it is shown
how a DP base generated post-verbally can undergo “ne-cliticisation”; (10) c shows
that the same can be applied to the subject of “arrive” but not to the subject of
“telephone” in (10) e:
(10) a.
b.
Ho
visto
tre
Have
seen
three
“I have seen three people”
persone
people
Ne
ho
viste tre
Of them have
seen three
“I have seen three of them”
t
i
i
13
c.
Sono
arrivate tre
Are
arrived three
“Three people have arrived”
persone
people
d.
Ne
sono arrivate tre
Of them are arrived three
“Three of them have arrived”
t
e.
Hanno telefonato
tre
Have
phoned
three
“Three people have phoned”
persone
people
f.
*Ne
hanno
telefonato
Of them have
phoned
“Three of them have phoned”
tre
three
i
i
i
t
i
Taking transitive verbs to have both an initial 1 (subject related) and initial 2
(direct object related) stratum - a reformulation of linguistic level (in a way, Θroles; see Perlmutter (1980: 198-200)) - he divides the intransitives into two
groups: the unaccusatives, which contain only an initial 2 stratum, and the
unergatives, which contain only an initial 1 stratum. He also claims (1980: 210)
that the selection of the auxiliary is determined by the type of initial stratum the
predicate contains. Initial 2 stratum will trigger selection of E aux, initial 2 A aux.
Burzio (1986) suggests a unified account of auxiliary selection and PtPPL agr: his
account, based on SI, rests on the existence of empty categories at S-S which create
a chain, responsible for triggering PtPPL agr. E aux selection is formulated as
follows (1986: 55):
(11)
“... ESSERE ASSIGNMENT: The auxiliary will be realized as essere
whenever a binding relationship exists between the subject and a nominal
contiguous to the verb...”
A “nominal contiguous to the verb” is defined (1986: 56) as a “nominal which is
either part of the verb morphology, i.e. a clitic, or a direct object. A direct object is
14
an NP in an A-position governed by the verb...”
This is translated into two configurations of binding relationships between:
i) an NP and its trace;
ii) the subject of the predicate and a clitic.
(12) a.
NP V
NP
b.
NP cl-V ...
(12) a
represents all the cases of NP-movement from object to subject position
- passives and ergatives ( in Burzio’s sense);
(12) b
accounts for E selection in all cases where the clitic SI is involved, i.e.
reflexive, reciprocal and impersonal constructions.
PtPPL agr is accounted for in the following way:
(13)
“...PAST PARTICIPLE AGREEMENT: A past participle will agree (in gender
and number) with an element holding a binding relationship with its direct
object...”
PtPPL agr is triggered, under this definition, in two configurations of binding
relationships between:
i) a subject NP and its trace;
ii) an object NP and its trace.
(14) a.
(14) a
(14) b
NP V
NP
b.
...
cl-V
NP
represents instances of unaccusative and passive constructions;
accounts for direct object clitics.
Given that both configurations, (12) a and (14) b partly overlap, the combination of
the two will account for all those structures in which E aux selection and PtPPL
agreement co-occur, as in reflexive constructions. The syntactic configuration in
(16) corresponds to sentences like (15) where the functional subject is linked,
through an intervening clitic to the canonical subject position:
15
(15) [Maria e
Giada] si
sono lavate (f, pl)
Mary
and Jade
themselves
are washed
“Mary and Jade have washed themselves”
i
(16)
i
t
i
NP cl-V NP
In all instances of movement it is assumed that the trace left behind by the pvDP
shares grammatical features of person, number and gender with the moved DP, as
stated in Chomsky (1981: 323):
(17)
“... If (α,
features
β) is a link of some chain C, α an argument, then β bears the same
as
α
...”
.
Although Burzio’s analysis can consistently explain a considerable amount of data,
by considering A selection and PtPPL agr between a PtPPL and a pvDP in situ in
complementary distribution, it could not explain the agreement shown in examples
(3) and (6) from Friulian 5.
In order to strengthen Burzio’s proposal, in Paoli (1996) I interpreted “ a nominal
contiguous to the verb”as an element carrying [+N] features, and identified it with
AgrO. Because of its pronominal nature, AgrO was able to bind the direct object
(DO) of the PtPPL. Furthermore, the “strength” of this node was considered
responsible for the difference between SI and Friulian: namely, the latter was
specified for “strong” features which had to be spelt out on the PtPPL, while the
former always caused the PtPPL to surface as the unmarked [m, sg] form unless
the DO underwent cliticisation 6.
Burzio’s (1986) account fails to capture data from a number of dialects (Cf
Brown (1988)). One of the many, Southern Lazio, displays lack of PtPPL agr with
the E aux. Rohlfs (1969:116) reports some examples: semo cercato - lit: we are
looked for (UM), semo vennuto - we are come (UM).
5
This difference in strength between the SI and the Friulian AgrO node is
captured by Cocchi (1994b: 89, note 5) in a different way. She assumes a difference
in timing in raising to the node: in Friulian the DO would raise to [Spec, AgrO]
before spell out; in SI this would not happen until LF. She further suggests that
this claim can find some support in Latin where in all transitive constructions
PtPPL agr with the DO was instantiated. In the transition to the Modern Romance
Languages this process has become non-overt, in that in all these languages the DO
raises to [Spec, AgrO] at LF, leaving the PtPPL in its [m, sg] unmarked form.
Medieval Romance represents an intermediate stage, where an alternation between
16
6
The structure obtained for (3), repeated here for convenience, is (17).
(3) Marie
e
a mangiadis
Mary
SCL has eaten (f, pl)
“Mary has eaten the sweets”
(17)
li caramelis
the sweets (f, pl)
AgrS
Spec
AgrS’
Marie
Agr
TP
e
a Spec
T’
[III, sg]
t
T
AuxP
t
Spec
Aux’
t
Aux
AgrOP
t
Spec
i
j
i
j
i
j
AgrO
AgrO’
VP
[f, pl] Spec V’
t
V
DP
mangiadis li caramelis [f, pl]
i
This solution has a number of theoretical and empirical problems which will be
dealt with in section 1.4.
1.4 -
Problems with Paoli (1996)
Although the analysis outlined in Paoli (1996) could provide a reasonably
the unmarked non-agreeing form and the agreeing form of the PtPPL can be seen.
Friulian, with all its varieties, would find itself in this intermediate position.
Although plausible, this would still not take into account the different feature
specification of the two nodes illustrated in (21) below.
17
consistent account of both SI and Friulian data, it leaves some questions
unanswered, illustrated below in (18).
(18) a. An issue arising from the analysis suggested in the previous section, but
primarily with Burzio’s analysis, is pointed out in Cocchi (1994b). She draws
attention to reflexive constructions, where the clitic which represents an indirect
object (IO) should not trigger PtPPL agr - remember that according to Burzio’s
analysis only a DO can trigger PtPPL agr. The examples are taken from SI:
(19) a.
Maria
si é
comprata
dei
Mary (f, sg)
refl. is bought (f, sg) some
“Mary has bought herself some books”
libri
books (m, pl)
b.
?Maria
si é
comprati
dei
Maria (f, sg) refl. is bought (m, pl) some
“Mary has bought herself some books”.
c.
Maria
se li
é
comprata / comprati t
Mary
refl. them (m, pl) is bought (f, sg) / (m, pl)
“Mary has bought them for herself”
i
libri
books (m, pl)
i
In Friulian the situation is slightly different, as shown in examples (5) a and b, but
it suggests even more strongly that an account based on Burzio’s analysis cannot be
accepted. Although (20) a is a reflexive construction, both auxiliaries can be
selected; furthermore, when the DO undergoes cliticisation, then only the A aux is
selected, and the PtPPL surfaces in its UM form.
(20) a.
b.
Marie
si a
comprât/ e
Mary refl. has bought(UM)/ is
“Mary has bought herself a dress”
Si
lu a comprât
Refl.
it has bought (UM)
“She has bought it for herself”
i
18
comprade un vistît
bought (f, sg) a dress (m, sg)
/*e comprade t
/is bought (f, sg)
i
Similar comments can be made about those verbs which allow both auxiliaries but
the choice reflects a semantic difference (Cf (1) b and c) and which cannot be
captured under Burzio’s analysis, as it stands.
More problematic evidence can be found in impersonal constructions, which show
that auxiliary selection and PtPPL agr are not two different sides of the same coin,
and do not, therefore, necessarily co-occur.
Examples (20) c, d, e and f, compare two intransitive constructions and their
corresponding si-impersonal counterparts. Only those intransitive verbs that
usually select the E aux display agreement on the PtPPL, which proves that
impersonal constructions do not show PtPPL agr by default:
(20) c.
Mario
ha camminato
Mario
has walked
“Mario has walked a lot”
molto
a-lot
d.
Si
é
camminato/ *camminati
SI
is walked (UM)/ *(m, pl)
“One has walked a lot”
e.
Mario
é
andato
Mario
is gone (UM)
“Mario has gone shopping”
f.
Si
a
to
é
andati/
*andato
SI
is gone (m, pl)/*(UM)
“One has gone shopping”
molto
a-lot
fare la spesa
do the shopping
a
to
fare la spesa
do the shopping
Although this matter will not be pursued any further here - see the solution
suggested in Cocchi (1994b) - it is indicative of the limitations of the conclusions
reached in Burzio (1986) and, consequently, in Paoli (1996).
(18) b. If we assume the functional subject of the PtPPL to be base-generated
in [Spec, VP] 7, and then raised to [Spec, AgrSP], we have to explain why AgrO is
7
I will follow Koopman and Sportiche (1988) in assuming that subjects of
19
not specified for the features carried by the subject. In fact in its raising by Spec-toSpec the subject moves out of [Spec, VP] and raises into [Spec, AgrO] in order to
reach its final position.
One way round this problem would be to say that AgrO is actually a functional
projection available only to internal arguments of the verb to raise into. This would
imply that if the syntactic subject did raise into it and transmitted its features to it
then there would be a clash in semantic features of some sort - something on the
line of “subject-ness” vs “object-ness”. Therefore it could be claimed that although
the structural subject does indeed raise into [Spec, AgrO] it does not transmit its
features to it. Alternatively, it could be assumed that AgrO does not project a
Specifier position as SS, but only at LF in order to allow for the object to check its
features.
Although possible, this solution relies on the assumption that AgrO is a projection
closely linked to the logical function of object rather than to its syntactic position,
a question addressed in point c below.
(18) c. Another problem which will be fundamental to the line of thought
presented in this dissertation arises from the fact that in Friulian, as well as in
Italian, agreement between a verbal form and its DO takes place only with
compound tenses formed with the PtPPL. This would actually suggest treating
PtPPL agr as a phenomenon strictly linked to the nature of the PtPPL rather than
to the function of DOs in general.
(18) d. A fourth problematic fact to take into consideration is the different
feature specification of the two Agr nodes:
(21)
AgrS [person, number];
AgrO [gender, number].
[person, number] are the features generally involved in subject-verb agreement in
SI and Friulian and other Romance languages. The features [gender, number], on
transitive and intransitive verbs are not base generated externally to the VP (as
assumed by Williams (1981)), but rather internally.
20
the other hand, are involved in the agreement instantiated between a noun and an
adjective modifying it. The PtPPL is a verbal form, so that both nodes should be
associated to verbal agreement. If this is the case, this difference in feature
specification would represent a problem for any attempts to give a unified account
of the two.
Kayne (1989) gives a parallel account for subject-verb and verb-object agreement,
but distinguishes between the two by claiming that while the agreement is direct in
the latter, it is mediated by an empty category. This is situated in a position
adjoined to IP - AgrP strictly speaking - in constructions with Wh-PtPPL agr, and
in [Spec, AgrP] with clitic-PtPPL constructions 8.
(18) e. The structure presented in (17) has a technical problem regarding
adjunction. The SCL is taken to be base generated under AgrS, so that when the
verb raises to this position it right-adjoins to the SCL. Kayne (1994) rules out rightThe agreement relation is expressed in terms of co-indexation : the Agr node
and the pvDP are co-indexed. Given that co-indexation is subject to locality
conditions, the fact that the auxiliary agrees with the subject and the PtPPL with
the pv DP could be easily derived: the verb will agree with its own subject but will
not be able to do so with the subject of the next verb higher up. The postulation of
an intervening empty category between the moved DP and the Agr node allows for
a more strict locality condition than a direct agreement between PtPPL and DP
would do, accounting for the ungrammaticality of (22) with the agreeing PtPPL:
(22) Paul a ...AGR [VP repeint
/ (*-es) les chaises]
Paul has
re-painted (UM)/ *(f, pl) the chairs (f, pl)
“Paul has repainted the chairs”
(22) suggests that a government relation has to hold between AGR and the NP and
no barriers - in Chomsky’s (1989: 87) sense - can intervene between the two in
order for agreement to be triggered. This is more forcefully illustrated in some
examples where Wh-movement has occurred (Kayne 1989: 86):
(23) a. Je me demande combien
de tables Paul
a
I
refl ask
how-many
of tables Paul
has
repeintes
repainted(f, pl)
“I wonder how many tables Paul has repainted”.
b. ... [combien de tables] [IP aux Paul [VP a [IP PtPPL AGR repeintes [e] ]]]
Were the empty category before the AGR absent, we would have to assume that
AGR agrees directly with the moved Wh-phrase across three maximal projections represented in bold in (23) b. By postulating the existence of that empty category
the locality constraints on the agreement process are not violated. Agreement will
be triggered when the DP moves to - or through - a position governed by AGR. In
Wh-PtPPL constructions this position would be adjoined to AGRP, in clitic-PtPPL
constructions it would be in [Spec, AGRP] as shown, respectively, by (24) a and b:
(24) a. ...[combien de tables] Paul a [IP [e] [IP AGR repeintes [e] ]]
b. Paul les a [e] AGR repeintes [e]
8
i
i
i
i
i
i
21
i
i
i
adjunction. The question of the position occupied by SCLs will not be pursued any
further here. The reader is referred to Poletto (1992) for a solution 9.
(18) f. The notion of “strength” invoked in Paoli (1996) in order to explain the
difference between Friulian and SI is rather curious. It is an intuitive concept and
cannot really be identified with the meaning “strength” has in the Minimalist
theory or others.
(18) g. Finally, SCLs and PtPPL agr are not two phenomena intertwined with
each other, given that there are Dialects which have SCLs but do not present PtPPL
agreement (e.g. Trentino).
Of the problems outlined above, (18) c and d are the most important ones.
Combined together, they suggest a similarity between PtPPLs and APs. The next
section will investigate further this suggestion, presenting Lehmann’s (1988) ideas
on the existence of two types of agreement, and, in a subsection, Giorgi and
Pianesi’s (1991) claim that the PtPPL is a verbal form with a “defective” nature
lacking a “speech time” reference.
1.5 -
Lehmann’s (1988) internal vs external agreement
Lehmann’s (1988) approach to agreement is functional. His basic thesis is that
Investigating the nature and function of SCLs from all Northern Italian
Dialects, Poletto comes to the conclusion that SCLs vary from dialect to dialect, and
accordingly, they occupy different positions. The CP is split in four different
projections and is referred to as the “Agreement field”. Inside it, each position is
associated with a syntactic feature and hosts a different type of SCL.
9
22
agreement is referential in nature and its function is to identify or re-identify
referents. Within this role he recognises two different types which he labels
INTERNAL and EXTERNAL. Internal agreement refers to noun phrase internal
agreement ( Lehmann’s concept of noun phrase is here taken to be a DP following
Abney (1987), Szabolcsi (1987), Fassi Fehri (1989)). All the elements involved in
internal agreement are c-referential, i.e. they all predicate over the same variable in
semantic expressions of the form
apple (x), red (x).
External agreement, on the other hand, is the agreement that is instantiated
between a DP outside the governing category of the agreement term and the term
itself, as in the subject-verb case.
As an example of internal agreement, Lehmann takes Latin constructions (1988:
56):
(25) illarum
duarum
bonarum
that.gen 10.pl.fem
two.gen.pl.fem good.gen.pl.fem
“Of those two good women”
feminarum
woman.gen.pl.fem
The example in (25) shows the agreement of determiner, numeral, attributive
adjective and noun. Lehmann observes that internal agreement may be specified but does not need to - for [case] but never for [person] features - more support for
this claim is found in examples of agreement between possessor and the possessed
item in Dyirbal.
Investigating external agreement in Abkhaz, Lehmann comes to the conclusion
that this type of agreement may be specified for [person] but never for [case]
features; this points to the fact that agreement specified for [case] is in
complementary distribution with the agreement specified for [person]. This claim
will not be taken any further here, since Lehmann interprets this difference as a
reflection of a difference in the function performed by the two agreement types,
internal and external. In particular , both are seen as (1988: 64)
10
“gen” here stands for Genitive Case.
23
“... successful ways of signalling a dependency relation, but external agreement is
more successful on purely syntactic relations, while Case marking is more successful
”
on semantic relations...
Although there is no direct reference to the agreement specified for [gender] features, it
is very significant that [person] features are never involved in instances of internal
agreement 11. This crucially suggests that PtPPL agr should be treated the same way as
DP internal agreement. The next subsection presents a formal representation of the
PtPPL’s nature, which makes it very similar to adjectives.
This claim is challenged by Hungarian, where DP internal agreement involves
[person] features. What is suggested here maintains its value and strength if Szabolcsi
(1981, 1994, among other works) is right in her analysis that sees the DP with a sentencelike structure, containing an inflectional node and being headed by a determiner. She
further claims that the argument frame of complex event nominals is identical to that of
the underlying verb. Thus, there is a close parallel between the syntactic structure she
gives to the DP and the one Chomsky (1986) gives to the CP. Compare:
Szabolcsi
Chomsky
DP
CP
Spec
D’
Spec
C’
D
(N+I)P
C
IP
DP
(N+I)’
NP
I’
DetP
(N+I)
I
VP
[±poss]
[±tense]
[(AGR)]
[(AGR)]
11
24
1.5.1 - [gender, number] features: Giorgi and Pianesi (1991)
Giorgi & Pianesi (1991) make clear reference to the nature of [gender, number]
features. Analysing the differences between Latin and SI compound tenses, they
adopt a revised version of Reichenbach’s (1966) theory of tense. They assign a
structure to compound tenses in which two sets of functional nodes are present:
one for the auxiliary and one for the PtPPL. The feature specification carried by the
two nodes directly reflect the categorial specification of the two verbs: the auxiliary
is specified for [person, number] and is a [+V, -N] category; the PtPPL is specified
for [gender, number] and is a [+V,+N] category.
Furthermore the two sets of nodes carry different semantic information. In the
syntax of temporal representation there are three primitive concepts: S, speech
time, denotes the time at which the sentence is uttered; E, event time, denotes the
time of the event e expressed by the verb; R, reference time, adopted to account for
the semantics of perfect tenses. In (26) an example (Giorgi & Pianesi (1991: 190)) is
given to exemplify this:
(26) When Mary entered the room at 5, John had already left at 4.
E R S
|___ |___
at 4 at 5
In (26) “John had already left the room” is the event and it precedes R, also
connected to an event, in this case Mary’s arrival. The pluperfect expresses a
consecutio temporis in which both E and R precede S.
Giorgi & Pianesi claim that morphemes expressing temporal relations (i.e. past,
future) and temporal representations (i.e. T-relations such as “S follows R”) are in
biunique correspondence (Cf 1991: 191).One of the consequences of this is that the
different tenses correspond to different temporal relations, and, more specifically,
that every morpheme corresponds to a certain T-relation and vice versa. Therefore,
given the T-relation S_R (S precedes R) the morpheme corresponding to it will be
the one expressing future tense (1991: 192).
This information is encoded in the two sets of inflectional nodes, one correlated to
each verbal form. For the present perfect they assume two nodes, T1 and Agr1
governing the auxiliary, T2 and Agr2 the PtPPL. T1 and T2 lexicalise tense
relations, S/R and R/E respectively.
25
The necessity of having two agreement nodes stems from the “incompatibility”
between T1 and T2's feature specification: T2 is “adjectival”, i.e. a [+V,+N]
category, T1 is “verbal”, i.e a [+V,-N] category. [person] features are here
considered typical of verbal categories, [gender] typical of adjectival categories.
In (27) an exemplifying tree (1991: 193) shows the auxiliary carrying S/ R
information, while R/ E are expressed by the PtPPL 12; the event e under both Vs is
co-indexed to show identity of event:
(27)
AGR1"
AGR1'
AGR1
T1"
T1'
T1
V"
S/R
[person,
number]
V'
V
AGR2"
e
AGR2'
i
AGR2
T2"
T2'
T2
V"
R/E
[gender,
number]
V'
V
...
e
i
The crucial points to extrapolate here are that
It is interesting to notice that R was required when these tense relations were
first introduced by Jespersen (1924) in order to account satisfactorily for the
semantics of perfect tenses. It is claimed that R is an indication of “perfectivity” of
the action.
12
26
i) the feature specification of PtPPL agr is due to PtPPL’s “adjectival” nature13; ii)
PtPPL agr, being specified for [gender, number] seems to be a phenomenon related
to the PtPPL rather than to the DO.
The similarity between PtPPLs and adjectives is shown in (28). DP internal
agreement in those languages where nouns are inherently defined for gender is
specified for the same features, [gender, number] (28) a and c are from SI, (28) b
and d from Friulian.
(28) a.
b.
La
The (f, sg)
bella
casa
nice (f, sg) house (f, sg)
SI
Le
biele
F
ciase
c.
Il
The (m, sg)
gatto
morbido
cat (m, sg) soft (m, sg)
SI
d.
El
giat
F
morbid
The PtPPL has, on the other hand, properties clearly typical of verb forms: it can
assign Θ-roles and Accusative Case. This is evident in absolute constructions,
where only DPs generated post-verbally can appear: presumably a subject DP is
ungrammatical because it cannot be assigned Nominative Case (Cf Belletti (1981,
1990).
(29) a.
b.
Pagât
il cont, Toni l’ é
lât cjase
Paid
the bill, Toni he is gone home
“Having paid the bill, Toni went home”
*Toni
senât,
l’
é
lât cjase
I will leave here open the issue of whether SI PtPPLs are really specified for
[+V, +N]. See chapter 3 section 3.3 and chapter 4 for a development of this idea.
13
27
Toni
dined
he is gone home
“Having had dinner Toni, he went home”
In summary, PtPPLs display both features typical of adjectival categories, agreeing
in [gender, number], and features typical of verbal ones, assigning Θ roles and
Case. These two are very significant observations and will be at the centre of the
analysis put forward in chapter 3.
1.6 -
Conclusion
Summing up, in this chapter we have seen that the Friulian data cannot find a
proper account in a syntactic analysis such as Burzio’s (1986). Lehmann and Giorgi
& Pianesi provide some evidence for the treatment of PtPPL agr as a phenomenon
directly related to the nature of the PtPPL, rather than to the general function of
DO - which is what the functional projection AgrO implies. In particular, the
similarity between PtPPL agr and DP internal agreement, suggests a semantic
affinity between PtPPLs and APs. Chapter 2 investigates the idea that the PtPPL is
a “verbalised adjective” which describes a state brought about in the pvDP. Some
semantic accounts of auxiliary selection and PtPPL agr ( Parisi (1976), Centineo
(1996)) and a summary of the relevant points are presented. It is shown, however,
that even a revised version of Centineo’s analysis does not adequately explain
Friulian data, and that a purely semantic account of PtPPL agr is not satisfactory.
This gives support to the idea that PtPPL agr is a syntactic phenomenon and it is in
syntactic terms that its mechanism can be expressed.
28
Chapter 2
Past Participle Agreement: Semantic
Accounts
2.0
2.1
2.2
2.3
2.4
2.5
2.6
Introduction
Parisi (1976)
Centineo (1996)
Problems with Centineo (1996): perception verbs and indefinite objects
Krifka (1989) and Tenny (1992): a solution to Centineo’s problems
What about Friulian?
Conclusion
2.0 -
Introduction
Chapter 1 has highlighted the affinity between PtPPLs and APs, suggesting that the
particular syntactic feature specification involved in PtPPL agr could be due to this
similarity between the two categories. More precisely, that PtPPL agr is due to the
PtPPL’s property of describing a state brought about in the pvDP. This chapter will
investigate a number of semantic accounts of PtPPL agr and reach the conclusion
that, although intuitively promising, a purely semantic analysis cannot account for
all the data. This is an important step forward in the understanding of PtPPL agr
and agreement in general.
Section 2.1 presents Parisi’s (1976) speculations. Parisi is the first to introduce the
concept of verbs which “initiate a state in one of its arguments” and to distinguish
29
between these and those which are not initiators of a state. To each predicate,
Parisi associates a logical representation in which the predicate is decomposed into
its basic unit of meaning. It is in these logical representations that the affinity
between adjectives and PtPPLs is evident.
Centineo’s analysis stems from Parisi’s, and reviewing Dowty’s (1979) Montague
Grammar and Foley & van Valin’s (1984) - Role and Reference Grammar, Centineo
formulates her own theory of auxiliary selection and PtPPL agr. Centineo’s
framework is discussed in section 2.2, together with some key notions such as
pivot, actor and undergoer from Foley & van Valin. Centineo’s analysis of auxiliary
selection is presented only for its relevance to the “theory of markedness” that she
applies to PtPPL agr.
Section 2.3 presents some speculations on the system created by Centineo and
discusses problems arising from SI perception and activity verbs.
To these a solution is suggested in section 2.4 where Krifka’s (1989, 1992) and
Tenny’s (1992) aspectual analyses of events are illustrated.
Finally, section 2.5 shows how, even under a revised semantic analysis, Friulian
data cannot be satisfactorily accounted for. Nevertheless, the observation that
whenever PtPPL agr is instantiated it is always with the PtPPL’s internal argument
will be very important in chapter 3 when defining the operational domain of the
agreement process. This will allow us to restrict it to the nodes c-commanded by
the PtPPL.
2.1 -
Parisi (1976)
An account of auxiliary selection in Italian, based on the semantic properties of the
verb, has been the focus of studies dating back to 1954 (Leone) and was carried out
throughout the Seventies (Leone 1970, Parisi 1976).
30
Parisi starts from the observation that the PtPPL form of the verb always expresses
something having been accomplished. Verbs are divided into two groups with
respect to this property.
The first category (A) includes all those verbs which, whenever the
accomplishment of the action is realised, initiate a state in one of the verb’s
arguments. In this category we can find both transitive and intransitive verbs, such
as lavare - to wash- which implies a state in its logical DO and uscire - to go out which imply a state in its logical subject. As a result of the two actions, we will
have, respectively, a “washed” DO and a “gone out” subject.
The second category (B) includes all those intransitive verbs that express an
accomplishment of something which will not initiate a state in any of the verb’s
arguments. An example is dormire - to sleep, where the subject of the action will
not be in a “slept” state after having slept 14.
Giving a graphic representation of this generalisation, Parisi expresses the ongoing
action as a wavy line, the accomplishment as a straight vertical line and the state
resulting from it as a straight horizontal line:
(30)
Cat A:
≈≈≈≈≈|
Cat B:
≈≈≈≈≈|
Extending the generalisation, he observes that all those verbs which contain the
component CHANGE in their logical representation, typically denote a state in one
of their arguments. A few examples are given (Parisi 1976: 80) with the
corresponding logical representations. It must be stressed that I will not take the
structures represented in the part b of all the examples as syntactic structures.
They are logical representations of predicates.
This of course would be a logical possibility, but not for Parisi who interprets
the predication on a syntactic, not logical level.
14
31
(31) a.
b.
Lucia
é
uscita
Lucy
is gone-out
“Lucy has gone out”
PRED
ARG
PRED ARG
CHANGE
out
Lucy
where out is usually a predicate (as a consequence of going out the subject of the
action is out), and the verb uscire is actually seen as the transition between two
states, that of being in and that of being out.
(32) a.
Lucia
é
morta
Lucy
is died
“Lucy has died”
b.
PRED
ARG
PRED ARG
CHANGE
not alive Lucy
(33) a.
Mario
ha ucciso
Mario
has killed
“Mario has killed Lucy”
b.
PRED
ARG
CAUSE
Mario
Lucia
Lucy
PRED
CHANGE
ARG
ARG
PRED ARG
not alive Lucy
Here morta - dead, not alive - is a state and both morire - to die - and uccidere - to
kill - are verbs that indicate a state in one of their arguments. With kill a new
operator is introduced, CAUSE.
32
The adoption of a decompositional approach where all predicates are reduced to
"atomic" logical representations and linked to each other by means of logical
operators is highly desirable, particularly on the level of linguistic generalization.
Predicates can be derived from one another by adding operators such as CAUSE
and CHANGE (Cf section 2.3 (51) where the breakdown of predicates into their
basic semantic units is applied to the four verb classes states, activities,
accomplishments and achievements. Predicates are then transformed into one
another by using the operators CAUSE and BECOME ) 15. This avoids treating each
The development of this type of decomposition analysis of word meaning
finds its roots in the writings of Hjelmslev (1943) and Jakobson (1936), and is then
applied to verbs by Lakoff (1965) who suggested that sentences like
Floyd melted the glass
were derived from deep structures like
(Floyd caused (the glass melt)).
In the late Sixties there were currents of thought among linguists who believed in a
deep level of underlying syntactic structure representing the meaning of a sentence
in pure semantic terms.
The mapping from the deep to the surface level was explained by McCawley (1968)
in terms of the so-called "Predicate Raising", a transformation which attached an
operator to the next one higher up until the surface structure was obtained.
Fodor (1970) brought some convincing evidence that this was not the case.
Investigating the behaviour of “do so” and instrumental adverbials which can only
be shared by two DPs if the DP is the subject of the modified verb, Fodor came to
15
33
verbal entry as unique and not derivable in any way - especially semantically - from
other expressions such as adjectives (e.g dead or not alive with to die) or other
verbs (e.g to kill with to die).
Under Parisi’s approach PtPPL agr is a phenomenon crucially linked with the
particular category to which a verb belongs. Agr with an argument of the verb will
be triggered by the PtPPL of those verbs which belong to category A, i.e. those verbs
which predicate of a state in one of their arguments, while it will not be possible
with the verbs belonging to category B.
With respect to SI, this rule presents subject - object asymmetries, in that PtPPL
agr is seen as compulsory when the state is predicated of the logical subject, while
it is optional when it is the logical DO in situ of which it is being predicated (Cf
examples (3) and (6) a and b from Friulian). The instantiation of this latter is
considered “substandard”, and Parisi formulates a rule which specifies that the
non-agreeing form of the PtPPL - in its unmarked masculine singular form - will be
preferred to the agreeing one with a DO in situ (1976: 82):
(34) a.
b.
I
soldati
The soldiers
? I soldati
The soldiers
hanno
have
vs
hanno
have
ucciso
killed (UM)
uccisa
killed (f, sg)
Maria
Mary (f, sg)
Maria
Mary (f, sg)
the conclusion that a process of decomposition and derivation by lexicalisation
cannot be accepted for “melt” and “kill” sentences.
As already pointed out, the structures presented here and those which will be given
in section 2.3 are not taken to be syntactic structures, as they were originally meant
to be, but just as logical representations of the semantics of predicates.
34
According to Parisi, when choosing between the two forms of the PtPPL illustrated
in (34) a and (34) b the deciding factor seems to be
“... the emphasis on the accomplishment of something... (1976: 98)”,
the soldiers’ activity in (34) a or the resulting state in (34) a, Mary’s death.
Perhaps in the Seventies people were still using sentences like (34) b rather
productively in SI; the same observation could not be maintained for today SI
where the semantic difference seems to have been lost.
Parisi draws a parallel between PtPPLs and adjectives on the following lines:
•
PtPPLs always express the accomplishment of something, and, if belonging to
category A, also predicate a state in one of their arguments;
•
adjectives always and only express a state in one of their arguments.
Graphically:
(35)
PtPPLs
≈≈≈≈≈|
Adjs
The parallel is clear also in the logical representation (Parisi (1976: 84)):
(36) a.
Maria
é
invecchiata
Mary
is got old
“Mary has aged”
b.
PRED
PRED
CHANGE
ARG
ARG
old
(37) a.
Maria
é
Mary
is
“Mary is old”
b.
PRED
ARG
old
Mary
Mary
vecchia
old
35
Parisi speculates on these similarities, and formulates a rule for auxiliary selection,
which is not directly relevant to my discussion. However, it must be mentioned
that those verbs which admit both auxiliaries, like correre - to run -, piovere - to
rain - (Cf examples (1) b and c) are accounted for in terms of the semantic
difference that arises between the two choices (1976: 87). There is a clear link
between the auxiliary and the meaning of the sentence.
(38) a.
Maria
é
corsa
a Milano
Mary
is run (f, sg) to Milan
“Mary has rushed to Milan (i.e. the running is towards Milan)”
b.
Maria
ha corso
a Milano
Mary
has run (UM) in Milan
“Mary has run in Milan (i.e. the running is within Milan)”
In (38) a correre is a verb of category A; Maria has undergone a change of state:
she finds herself in Milan as a result of the action of running. In (38) b correre is a
verb of category B; Maria does not find herself in a state predicated by the verb as a
result of her action.
Centineo speculates on these differences and adopts concepts derived from Foley &
van Valin (1984), Vendler (1967) and Dowty (1979) to justify auxiliary selection and
PtPPL agr. Section 2.2 presents her analysis.
2.2 -
Centineo (1996)
Centineo (1996) reviews Parisi’s analysis and emphasises that subsuming under
the same logical expression both agreement of predicate adjectives and PtPPL agr
the distinction between the use of essere as copula and as auxiliary can be
eliminated, a conclusion underscored by Parisi 16.
This position is also independently supported by Benveniste (1966) and
Kayne (1993), where an analysis that sees have derived from be through the
36
16
Centineo’s theory of auxiliary selection and PtPPL agr in SI rests on some basic
concepts which she derives from Vendler (1967), Dowty (1979)’s Montague
Grammar and Foley & van Valin’s Role and Reference Grammar.
From Vendler (1967) and Dowty - but originally from a tradition finding its roots in
Aristotle's Metaphysics - she derives the classification of verbs 17 for the
explanation of auxiliary selection phenomena.
This classification does not create a biunique relation between a verb and a class.
As Foley & van Valin observe (1984: 39), there is a certain degree of mobility
between them. Activity verbs, for example, may become accomplishments if a
definite goal is added. An example is given by walk, an activity verb which turns
into an accomplishment when a goal is added, as in walk three miles. The reverse
is also true: an accomplishment verb can become an activity if its object is a mass
or generic noun. Compare eat a bag of popcorn and eat popcorn, where the former
is an accomplishment and the latter an activity.
Dowty (1979) lists a set of tests distinguishing between activity and
accomplishment verbs, one of which is the combination with specific adverbials.
Activity verbs can happily combine with durative adverbials of the type V for an
hour which can be equated by spend an hour V-ing; in these cases, V for an hour
entails V at all times in the hour, and x is V-ing entails x has V-ed (cf Dowty 1979:
56). Accomplishment verbs prefer time-span adverbials of the type V in an hour
and allow duration adverbials only marginally. We do not find the same
entailments: x is V-ing entails that x has not V-ed yet.
Under Dowty’s account the verbs selecting the A aux are all intransitive activity
verbs. Those verbs which select the E aux are, on the other hand, intransitive verbs
and the group is composed by predicates describing states or conditions of being,
locative predicates and some predicates of perception and possession. Applying the
logical structure of verbs to their choice of auxiliary and the system outlined in
Vincent (1982), Centineo comes to the conclusion that the selection of the auxiliary
incorporation of a functional head into the latter is extended to “have” and “be” as
auxiliaries.
Verbs are classified in 4 groups, states (e.g. to know), achievements, (e.g. to
recognise), activities (e.g. to drive) and accomplishments (e.g. to wash). See (48)
for their logical representation.
37
17
depends crucially on the semantics of the pivot. In her own words,
"... Verbs expressing an actor pivot occur with avere, while verbs expressing a
pivot which is semantically "affected", that is an undergoer, ... occur with
essere...(1996: 257)”18
.
The notions of pivot and actor and undergoer from Foley & van Valin (1984: 29 and
110 reported here in (39) and (40) respectively), who give a description of “subject
of a sentence” basing it on Heath (1975) and Dixon (1979) 19:
(39) “...
Provisionally we may characterize the
actor
as the argument of a
predicate which expresses the participant which performs, effects, instigates
or controls the situation denoted by the predicate, and the
undergoer as
the argument which expresses the participant which does not perform,
This claim finds more support in the unaccusative - in Burzio’s sense - verbs.
Their argument is not an agent. If we compare two verbs such as arrivare - arrive,
and telefonare - telephone, it is clear that arriving is not something you do, it’s
really something that happens to you, while telephoning is definitely a voluntary,
agent-driven action. The single argument of unaccusatives usually denotes an
entity that undergoes a change of state or location.
18
Actor and undergoer, the two arguments of a transitive predication, are
concepts inherited from Gruber (1965) and Fillmore (1968). The terms were
actually introduced by Foley & van Valin to capture the semantic functions of
subject and object: their main aim was to mark the distinction between actor and
syntactic subject and between undergoer and syntactic DO. The syntactic subject of
an action is not necessarily an actor, and in unaccusative verbs for example, the
subject can be the undergoer.
19
38
initiate or control any situation, but rather is affected by it in some way...
(40)
”
“...The pivot of a syntactic construction is the NP which is crucially involved
.
in it; i.e. it is the NP around which the construction is built...”
The adoption of these concepts is necessary for Centineo to formulate of “marked”
pivot choice. However, since these notions do not play any part in the analysis that
will be put forward in chapter 3, I will not pursue any specific objections to the
theoretical details of the notion of pivot. This is not to say that the notion is
without conceptual as well as empirical problems 20.
Centineo equates the notion of “surface syntactic subject” in English and in SI to
the notion of pivot, so that two main types of constructions are derived: the active,
or unmarked, where that actor of the sentence is the pivot, and the passive, or
marked, where the undergoer is the pivot.
Centineo creates a correspondence between this type of markedness and what she
defines as the "theory of marked pivot choice", i.e. the pivot’s degree of
“affectedness” in a sentence. Verbs are placed following a crescendo of "affected
state" of their external argument, where [-affected] is the least marked one and
[+affected] the most marked one. In this scale pivot-less verbs, such as impersonal
I will here point out just a major problem that the notion of pivot poses.
The definitions given in (39) and (40) are weak for their being vague in general
and consequently difficult to falsify. Considering double-object constructions like
(41) can help to clarify this point.
(41) a. MAX gave Peter a pen
b. Max gave PETER a pen
c. Max gave Peter a PEN
Capitals indicate stress. The sentences in (41) could be uttered as answers to,
respectively, the following questions:
a. Who gave Peter a pen?
b. To whom did Max give a pen?
c. What did Max give to Peter?
Each of the questions is aimed at eliciting different information, the agent of the
action, the goal and the theme. Each of these Θ-roles will be, in turn, the pivot of
the sentence, i.e. “the NP around which the construction is built”. All sentences in
(41) are syntactically equivalent, but each of them has a different pivot. Were pivot
a notion we could successfully apply to Syntax, we would expect sentences with
different pivots to be syntactically different.
Thus, the notion of pivot will be only referred to here because it is relevant to
Centineo’s analysis, surely not for its intrinsic strength.
20
39
verbs, are considered the most marked instance of the range.
To markedness of pivot choice corresponds markedness of auxiliary selection: the
E aux is selected by those verbs whose pivot is the affected argument of a stative
predicate in logical structure 21, the A aux will be selected by all the other verbs.
This concept of markedness is applied to PtPPL agr, too:
(42) a.
“... PtPPL agr is obligatory when the argument of the stative predicate
is the pivot of the construction, while it’s optional when the argument
of the state predicated is a non-pivot undergoer. ... In particular, the
optional PtPPL agr will occur only with the second argument of
transitive accomplishment, state and achievement verbs, while it will
not be possible in the case of transitive activity verbs...”
b.
(1996: 267).
"... the PtPPL agrees with the patient or theme argument of a state
predicate..."
(1996: 269).
In stating the “optionality” of PtPPL agr with a DO in situ Centineo is able to
account for the occurrence in SI of what has been called here a “learned” agreement
(Cf chapter 1, section 1.1) as opposed to the one spontaneously produced by
speakers of Friulian.
Section 2.3 discusses some problems arising from Centineo’s solution in dealing
with SI data.
“Affected pivot” here refers to the argument whose state or location or whose
undergoing a change in state or location is predicated of.
21
40
2.3 -
Problems with Centineo (1996): perception verbs
and indefinite objects
Looking more closely at the semantic theoretical framework on which Centineo’s
analysis rests, in particular at those categories of verbs which allow or require
PtPPL agr irrespective of the auxiliary they select, the following conclusion can be
drawn:
(43) in SI PtPPL agr is triggered by states, achievements and accomplishments by
both intransitive and transitive verbs.
Centineo clearly states (1996: 267) that the optional PtPPL agr occurs with the
second argument of transitive accomplishment, state and achievement verbs, but it
is not possible with transitive activity verbs. She derives some support from Italian,
where apparently (44) is not possible, even for those speakers who would admit the
use of a PtPPL agreeing with an in situ DO:
(44) *Maria ha mangiata
pizza
Mary
has eaten (f, sg)
pizza (f, sg)
“Mary has eaten pizza all evening”
tutta la
sera
all the evening
Why should this be so? What is the difference between a transitive activity verb
and, let’s say, an accomplishment verb?
The answer could be found by looking at the logical representation of verbs
belonging to these categories reported below from Foley & van Valin (1984: 39):
(45)
Verb class
Logical structure
a. State
predicate' (x)
b. Achievement
BECOME predicate' (x)
c. Activity
DO (x, [predicate' (x)])
d. Accomplishment
φ CAUSE ψ
(where φ is normally an activity verb and ψ an achievement verb)
The logical representation of accomplishment verbs could be rewritten as:
41
(46) [DO (x), [predicate' (x)]] CAUSE [BECOME predicate' (y)]
Achievement and accomplishment verbs contain in their logical representations
the atomic unit predicate’ (x), where (x) refers to the internal argument of the
predicate. In the representation of activity verbs, (x) can refer either to the external
argument - in intransitive activities - or to the internal one - in transitive activities.
It is difficult to see how , having a sentence like John runs, an intransitive activity,
the predicate run could predicate of a state of John’s; perhaps John would be in a
state of movement, but this seems vague and not specific to the verb under
consideration. I will leave the question open here.
We can draw a parallel between the classification in (45) and the one given by
Parisi (1976): state, achievement and accomplishment verbs correspond to those
verbs which Parisi includes in his category A verbs, i.e. those initiating a state
predicated by them in one of their arguments and which contain the abstract
operator CHANGE. The operator BECOME in (45) can be seen as containing the
idea of change from one state to another and by virtue of this be assimilated to
Parisi’s CHANGE.
PtPPL agr is instantiated by those verbs that initiate a state in their internal
argument and these verbs belong to the categories listed in (45) as a, b and d, and
all these verbs reduce to the same atomic unit predicate’ (x), where (x) is crucially
the verb’s internal argument. predicate’ (x) is also the logical representation
assigned to adjectives, so that the affinity between the two categories is formally
expressed.
Summing up all this in graphical terms, this is what we obtain:
(47) a. ____|
b. ____|___
c.
_______
• Intransitive activity verbs - No PtPPL agr;
•Transitive activity verbs, accomplishment
and achievement verbs - PtPPL agr;
• State verbs and adjectives - agr.
(47) combines Parisi’s and Centineo’s accounts. Some problems arise from it. The
42
condition put by Centineo on PtPPL agr is that it is only instantiated between the
PtPPL and the undergoer - i.e. the “affected” party in an action. Does this mean
that in (44) the pizza is not affected by Maria’s eating it? Surely this is not the case,
since after Maria’s action the pizza has undergone a change of state.
On a similar line of thought, perception verbs, which are classified under her
system as achievement verbs, come to represent a problem. In the sentence
(48) John has heard a noise
John is the actor and noise the undergoer. Again, the definition of actor and
undergoer given in (39) is proved not to be very appropriate: the idea of
"performer, effector, instigator or controller" entailed by the term actor, cannot
really apply to John, given that he has no control over his hearing coming about. A
parallel observation can be made for the idea of "affected" by an action implied by
the term undergoer: the role of undergoer cannot really be applied to the noise,
given that the noise is completely unaffected by John's hearing of it . In (48) it is
really the actor who undergoes a change of state, passing from a state of
unawareness of the noise to one of awareness of it, and not the noise 22.
Still, the following should be accepted in SI by whoever allows for PtPPL agr with
an in situ DO (Cf section 2.5 for an account of Friulian cases):
(49) Piero
ha sentita
Piero
has heard (f, sg)
“Piero has heard the story”
la storia
the story (f, sg)
Agreement takes place between the story and the PtPPL, even if, according to
Centineo, PtPPL agr is equal to affectedness. Piero in (49), who is the real entity in
which the change of state takes place, agrees with the auxiliary and not with the
Rogers (1971), classifying verbs under three headings, distinguishes two
groups of verbs, cognitive and active. Some examples are cognitive - see, hear,
feel; active - look at / watch, listen to, feel. He notices that the subject of the active
verb is intentionally doing something and is responsible for it; the subject of the
cognitive verb, on the other hand, can see or hear or feel something accidentally.
The two would also be differentiated syntactically: the active listen to would have
the same structure as hear but it would be embedded in a phrase headed by the
abstract operator DO. If this is correct, then it seems that the operator DO is
“responsible” for the notion of volition (and/or intention) on the part of the
subject.
22
43
PtPPL.
Since the notion of “change of state” and “affectedness” are the cardinal points on
which Centineo’s, but primarily Dowty’s , accounts are constructed, it would not be
desirable to discard them. On the other hand, the problems presented here are very
strong and cannot be ignored or treated as a kind of “exception”. Below it is shown
how Dowty’s revision of verb classification, adequately accounts for activity verbs,
but fails to capture agreement with perception verbs.
In Dowty’s (1979: 184) revision of verb classification verbs are categorised by their
temporal properties in the framework of interval semantics and by agency. In his
revision, accomplishments and achievements are collapsed in the same category,
reorganised into four groups depending on whether they imply a single or a
complex change of state and a non-agentive or agentive action. Those verbs
belonging to this new class can also be referred to as the “definite change of state”
predicates, and they differ from activity verbs defined as “indefinite change of
state”. The line between the two can be drawn following the syntactic test:
Does “X was V-ing (pragmatically) entail X has V-ed 23”?
Taking an activity verb like ballare - to dance - and an accomplishment predicate
like mangiare la torta - to eat the cake -, the test yields the desired result:
entails
(50) a. Maria stava ballando
“Maria was dancing”
b. Maria stava mangiando la torta
torta “Maria was eating the cake”
Maria ha ballato
“Maria has danced”
does not entail
Maria ha mangiato la
“Maria has eaten the cake”.
Although this is a plausible solution for activity verbs, perception verbs would still
be problematic given that achievements and accomplishments are collapsed into
23
Here I will take “pragmatically” to mean “ X has V-ed and is no longer V-ing”.
44
one category. Furthermore the concept of “definite” and “indefinite” change of
state is not very clearly defined and, consequently, not very efficiently applicable.
In summary, transitive activity verbs and perception verbs pose a problem for
Centineo’s conditions on PtPPL agr. Dowty’s revision of verb classification could
account for activity verbs but not for perception verbs. In the next section it is
shown how Krifka (1989, 1992) and Tenny (1992) can allow us to maintain
Centineo’s analysis with some modifications in the way of expressing the state of
the internal argument triggering the agreement in. This successfully accounts for SI
data. However, the modifications still fail to account for Friulian transitive activity
verbs, presented in section 2.5.
2.4 -
Krifka (1989) and Tenny (1992): a solution to
Centineo’s problems
The conditions put by Centineo on the “affected state” of the PtPPL’s internal
argument are problematic when dealing with activity and perception verbs. In
order to modify her analysis two important notions, predicates having an ending
point and the centrality of the internal argument of a verb, are investigated here.
It will be remembered from section 2.2, that one of the tests that makes possible a
distinction between activity verbs on the one hand and accomplishment and
achievement verbs on the other is the compatibility of a verb phrase with two time
adverbials, in an hour and for an hour. In-phrases seem to produce well formed
sentences only when modifying accomplishments, while activity verbs are only
compatible with for-phrases. The following examples illustrate this property:
(51) a. Mary has danced for an hour / *in an hour
b. Mary has eaten the cake in an hour / *for an hour.
When an accomplishment verb is followed by an indeterminate NP it turns into an
activity verb:
45
(52) Mary has eaten cakes for an hour / *in an hour.
Krifka (1989: 76) shows that durative adverbial expressions like for an hour select
for what he calls cumulative event predicates - using a term first introduced by
Quine (1960: 91), while time-span adverbials like in an hour select for quantized
ones. This distinction is carried over from the distinction originally made between
cumulative and quantized NPs. A quantized NP is an NP which denotes an object
with precise limits, i.e. an easily identifiable entity. A cumulative NP, on the other
hand, is an NP which denotes something without clear limitation, like mass nouns
or bare plurals. The contrast is clear with the quantized NP a pen and the
cumulative NP milk.
The parallel between predicates and NPs is made on the basis of similar differences
holding between cumulative and quantized NPs on the one hand and telic and
atelic expressions on the other. A verbal expression is atelic if its denotation has no
set terminal point; it is telic if it includes a terminal point. An example illustrating
this contrast is given by the verb “swim” in two different constructions, swim and
swim a mile. In the latter but not in the former the action has a terminal point
expressed by the reaching of the goal “a mile”.
Krifka (1989: 91) gives a formal definition of an event with a terminal point:
(53)
“... An event e of type
φ can be said to have a terminal point t relative to φ
iff it is the final temporal point of e and there are no e’ of type
φ with either
⊂Ee or e ⊂Ee’ (i.e. e’ proper part of e or vice versa) which have an earlier or
e’
a later final temporal point; otherwise, e has no terminal point relative to
φ
...”.
There seems to be a semantic parallel between objects with precise limits and
events with terminal points, and between objects without clear limitation and
events with no set terminal point. Telic verbal expressions are those identified by
Vendler (1967) as accomplishments, and atelic are activities. Typically, cumulative
expressions yield atelic verbal predicates, quantized telic ones.
Drawing from examples from the Slavic languages Krifka (1992) shows how
definiteness of the NP and perfectivity interpretation of the verb are intertwined.
46
Quantization is a component of the meaning of perfectivity, so that its definiteness
or lack of it can affect the temporal constitution of the verb, and the reverse is also
true. Aspectual properties seem to have some bearings on the “affectedness” of the
internal argument.
The idea that aspectual properties of verbal expressions play a fundamental role in
sentence interpretations is also found in Tenny (1992). In formulating the
Aspectual Interface Hypothesis - a rule constraining the mapping between syntax
and semantics - she claims that it is the aspectual structure associated with the
arguments of verbs that constrain "...the kinds of event participants that can
occupy these positions" (1992: 2) - she refers in particular to the characteristics
that elements base generated in object position must have in order to be able to
appear in that position24. It is only the aspectual part of the thematic structure
which is visible to and interpretable by the syntax.
The internal argument of a verb is seen as the carrier of the aspectuality by virtue of
which it "measures out the event over time" (1992: 3) - a similar view but
expressed in terms of space is found in van Voorst (1988: 27). It is therefore the
internal argument which delimits the event, and its definiteness is translated into
delimitation when interpreting the sentence.
Tenny further claims that a highly constrained mapping of semantics into syntax
can only be reached if based on aspectual properties rather than Θ-roles.
If we apply this notion of the event being delimited linguistically within the VP
through reference to the "measuring out" action of the internal argument, the
problem presented by perception verbs and the notion of affectedness can be
overcome. “Internal argument as object of termination” can replace “non-pivot
undergoer” in (42) in the account of all the instances of PtPPL agr . In particular, it
Tenny reaches the same advantage as Centineo has, i.e. the adoption of the
notion of pivot to avoid referring to Θ-roles, by claiming that the syntax can only
“see” certain syntactic/aspectual structures associated with the thematic roles, but
not the roles themselves.
24
47
could be added that the PtPPL must “describe” the state in which its internal
argument finds itself as a result of the action - property expressed by the relation
predicate’ (x). Thus, the “noise” will be in a “heard” state after having been heard
and will be the element “measuring out” the event. By virtue of these two combined
properties it will trigger PtPPL agr.
Turning now to transitive activity verbs, maintaining that a definite DO “measures
out” the event predicated by a verb, it is clear that not having a definite DO which
can delimit the event, an activity verb has no set terminal point. This is translated
on the syntactic level by the lack of PtPPL agr in sentences like (44). On the
semantic level, the DO and the predicate are to be considered a fusion into one unit
as far as meaning is concerned. Applying this to sentence (44), the following
analysis can be reached: eat is one predicate (V), pizza is another one (N), and
eating pizza is another predicate (V). It could be claimed that eating pizza belongs
to the set of the eating activities being a subset of it - eating pizza is more specific
than just eating. The focus is on the activity itself rather than on the resultant state
of the pizza after the action has taken place.
From a different perspective, if we extract a fraction of the period of time in which
the action of eating pizza is taking place, what we will have is a proper part of the
event which can still be considered as eating pizza, no matter how little that
fraction of time is - of course if the fraction of time selected is so little that the pizza
only gets bitten but not eaten, then it will not be a proper part of the event of eating
pizza.
If, on the other hand, we extract a little fraction of time during which the action of
eating the pizza takes place, that part of the event cannot be considered as a
proper part of an event of eating the pizza - the event will be such only when the
pizza has been eaten completely.
Tenny’s notion of DO as measuring out the action combined with Krifka’s idea of a
NP’s definiteness being intertwined with the perfectivity interpretation of the verb,
provide a successful way of modifying Centineo’s semantic account of PtPPL agr.
This could then be linked to the syntax by assuming - as Tenny (1992) does - that
there are strong correspondences between meaning and syntactic structure,
48
especially between logical form and syntactic structure. The Aspectual Interface
Hypothesis (Tenny 1992: 2) clearly states that the mapping between the two is
governed by aspectual properties and that only these are visible to the syntax.
Finally, the PtPPL’s aspectual properties as the main verb of a “measured out”
event are syntactically expressed by the instantiation of PtPPL agr.
Although the theory developed so far successfully accounts for SI, Friulian
transitive activity verbs still present a problem: the next section illustrates how.
2.5 -
What about Friulian?
Turning now to Friulian data and applying to it the verb classification, (54) what
we obtain is - a stative, b and c accomplishments, d achievement :
(54) a.
Marie
e
a
savude le rispueste
Mary SCL has known (f, sg) the answer (f, sg)
“Mary has known the answer”
b.
E ai lavâs
i
plas
SCL have washed (m, pl)
the dishes (m, pl)
“I have done the washing up”
c.
E ai
mangiâs
i
spaghes
cul sugo
SCL have eaten (m, pl) the spaghetti (m, pl) with-the sauce
“I have eaten spaghetti with the sauce”
d.
Marie e a
ricognusude
Mary SCL
has recognised (f, sg)
“Mary has recognised your grandma”
to none
your grandma (f, sg)
PtPPL is instantiated in all of the above categories, which is what we would expect
following Centineo’s modified analysis. The revised version of the semantic
49
account could also deal with perception verbs where PtPPL agr s triggered:
(55) a.
sg)
b.
Piero
el a sintude
Piero
SCL has heard (f, sg)
"Piero has heard an interesting story"
une storie
a story
interesante
interesting(f,
Atu
viudude
to sûr?
Have-you seen (f, sg)
your sister (f, sg)
"Have you seen your sister?
Some serious problems arise when considering activity verbs. In (56) a and b
PtPPL agr is triggered by a transitive activity verb:
(56) a.
b.
E ai mangiadis
cussutis
dute
SCL have eaten (f, pl)
courgettes (f, pl) all
“I have eaten courgettes the whole day”
le dì
the day
E ai cues
pes
di poles dute le dì
SCL have cooked(m, pl) breasts (m, pl) of chicken all the day
“I have cooked chicken breasts all day long yesterday”.
Here we have a bare NP - cussutis, courgettes and pes di poles, chicken breasts
respectively, which, because of their being indefinite, should not be able to
“measure out” the event of eating and cooking and consequently should not cause
the instantiation of PtPPL agr. This clearly is not the case as can be seen from the
examples: PtPPL agr is triggered in both cases.
More problems arise with meteorological verbs. In SI they can select both A and E
auxiliaries, so that both forms are possible:
(57) a.
É piovuto
Is rained
"It has rained"
b.
50
Ha piovuto
Has rained
"It has rained"
There is a difference in interpretation between the two which is captured by
Centineo’s analysis. When the focus of the sentence is placed on the activity of
raining, only the one with the A aux is possible. On the other hand, if the stress is
put on the quantity of water that has fallen, only the E aux is selected and PtPPL
agr is triggered:
(58) a.
b.
Ieri
sono piovuti
20 cc
Yesterday are rained (m, pl) 20 cc
"Yesterday 20 cc of water fell"
di
of
acqua
water
Ieri
ha piovuto
tutto il giorno
Yesterday haverained (UM) all the day
"Yesterday it rained the whole day".
In (58) a piovere is an intransitive verb, whose only argument cannot really be
affected by the action of which it is only formally a subject. In (58) b piovere is still
an intransitive verb but this time its subject is a referential NP, a quantity of water.
It could be claimed that this quantity of water is affected by the “raining” action - it
moves from one position to another - and agreement between this latter and the
PtPPL is triggered, as we would expect under Cenineo’s analysis. Accordingly,
while the emphasis in (58) a is on the action of raining, in (58) b it is on the
quantity of water fallen as a result of the raining.
The same does not apply to Friulian, which does not allow for the alternation of
auxiliaries with meteorological verbs: only the A aux is selected, irrespective of
whether the emphasis of the sentence is placed on the action or on the quantity of
water that has fallen. Even in the latter case, Friulian still selects the A aux and no
agreement is triggered:
(59) a.
Iar
l'
a
plot
Yesterday SCL has rained (UM)
“Yesterday it rained all day”
51
dute le dì
all the day
b.
Iar
l’
a plot
vinc
sentimetros di aghe
Yesterday SCL has rained twenty centimetres of water
“Yesterday there came down 20 centimetres of water”
In Friulian the E aux is not selected: (56) and (59) put forward some evidence
which is unequivocally showing that the semantic theory towards which we have
gradually built does not satisfactorily account for Friulian. This is a consistent
piece of evidence that discards the idea that PtPPL agr is regulated by semantic
factors related to aspectuality and verbs classification.
At this stage it is necessary that we review the situation to see what has been
achieved in this chapter.
2.6 -
Conclusion
Clearly a semantic account formulated in terms of aspectuality does not adequately
explain the Friulian data. The initial observations which motivated a reconsideration of the phenomenon PtPPL agr have not found an appropriate
interpretation. Briefly I will recapitulate what they are:
(60) a. AgrS and AgrO are specified for different features, respectively: [person,
number] and [gender, number];
b. Agreement between a verbal form and a full DO is instantiated in SI and
Friulian only with the PtPPL, suggesting the phenomenon is directly dependent on
the PtPPL’s nature;
c. The agreement instantiated between a noun and an adjective is specified for
the same features carried by the PtPPL agreeing with a pvDP, i.e. [gender,
number];
d. Finally an observation - derived from Parisi (1976) and Vendler’s (1967)
verbal classification - about the nature and function of the PtPPL as an element
“describing the state” of its internal argument - with the exceptions noticed for
cognitive and perception verbs.
52
Although the above points have not found an adequate interpretation, what this
chapter has shown is that PtPPL agr is triggered always with the internal argument
- this latter being in a particular relation to the verb, and that the PtPPL predicates
of its internal argument’s state. This captures the similarity between PtPPLs and
adjectives, although the two categories are not identical.
Both the affinity and difference between PtPPLs and APs is captured in the next
chapter, where a syntactic analysis of PtPPL based on a process of feature-copying
operational in a “proper domain” is presented. Under the new account, both PtPPL
and DP internal agr are instantiated by the same process of feature-copying; the
differences between PtPPLs and APs are captured in the different operational
domain in which the mechanism is effective.
In the next chapter I will look at the position occupied by the PtPPL with respect to
different adverbials, following Cinque (1994, forthcoming) and adapting the
structure for the Past Participial Phrase suggested by Belletti (1990, 1991). This will
provide us with some support for discarding the existence of a functional
projection responsible for PtPPL and DP internal agr, a step forward in the
direction of understanding agreement.
53
Chapter 3
Past Participle and adjectival agreement:
a unified account
3.0
3.1
3.2
3.3
3.4
3.5
Introduction
Adverb placement
Theoretical assumptions
The structure
Relevance of the different domains in DP internal and PtPPL agreement
Final remarks and Conclusion
3.0 -
Introduction
The last section of the previous chapter has shown how Friulian data casts
serious doubts upon an aspectual semantic account of PtPPL agr on the lines of
PtPPL agr taking place when the internal argument is the “affected” or “terminal”
point of the action. In spite of this fact, the semantic investigation has nevertheless
emphasised that PtPPL agr in Friulian is triggered always between the PtPPL and
its internal argument. This, together with the semantic affinity between adjectives
and PtPPLs, represents a very significant observation, and in this chapter I will
suggest a possible syntactic interpretation that could be given to them: the
mechanism responsible for adjectival and PtPPL agr is the same, but the syntactic
domains in which this mechanism operates are different.
Some empirical evidence against the existence of an AgrO node triggering
agreement comes from the results of tests carried out on the relative position of
54
different adverbials and the PtPPL in SI and in Friulian. If PtPPL agr were
triggered by the PtPPL raising into and AgrP, we would expect the PtPPL in
Friulian to be higher than in SI in constructions with a pvDP in situ. This is not
what we see in section 3.1, where we follow Cinque (1994, forthcoming) in
determining the order of the adverbials under consideration and draw significant
conclusions from his speculations about the obligatory movement of the PtPPL.
Section 3.2 presents the theoretical framework in which the present account is
formulated and the assumptions on which the analysis rests.
Section 3.3 illustrates how both the adjectival and PtPPL agr can be subsumed
under the same mechanism of feature copying (Postal (1964)), and stresses the
different domains in which feature copying performs its action in the two cases.
Finally, section 3.5 puts forward some speculations on the nature of the agreement
specified for [gender] versus the one involving [person] features. The
generalisations reached there hold for the Romance languages, but they need to be
tested cross-linguistically. This is left open for further research.
3.1 -
Adverb placement
Pollock (1989), whose work was inspired by Emonds (1978) and Klima (1964), has
provided some support for the idea that several syntactic phenomena connected
with word order can be interpreted in a very revealing way if taken to be an
indication- or at least a reflection - of verb movement processes.
We will adopt Pollock’s framework and methodology as a means of investigating
the movement properties of verbal elements. The strength of this interpretation lies
in the assumption that the position occupied by different classes of adverbials is
constant across languages (Cf Cinque (1994), forthcoming)25 .
This is by no means a universally accepted belief. Compare Williams (1994)
for a contrastive view. He rejects the idea that adverbs are distributed in the same
way across languages, and consequently that any alternation observed in their
position with respect to verbs must reflect verb movement. Iatridou (1990), too,
opposes the strategy of basing the individuation of the sequence of functional
categories on adverb positions.
25
55
The most significant and revealing results for this research can be obtained by
investigating those adverbials placed within the projections which delimit PtPPL
movement, as specified in Cinque (1994). The adverbials tested are mica - a reenforcing negative polarity adverb (since no literal translation can be given, it will
be referred to as “NegPol”), già - already, più - any more, sempre -always and
completamente - completely. Additionally, also the lowest adverbials, tutto everything and bene - well, will be tested, given that they seem to mark the
obligatory position to which the PtPPL must rise in SI.
The relative order of these adverbials in SI is given in (61):
(61) mica > già > più > sempre > completamente > tutto > bene (Cinque 1994:
170)
The Friulian counterparts of the above adverbials are, in the same order mingul,
zà, plui, simpri, completaminti, dut and ben. The examples in (62) show that they
occur in the same order in Friulian and SI:
(62) a.
Mi
an spiegât dut
ben
To-me have explained everything
well
“They have explained everything well to me”
b.
*Mi
To-me
an spiegât ben dut
have explained well everything
c.
E an rifat
completaminti dut
SCL have re-done completely
everything
“They have done again completely everything”
d.
*E an rifat
SCL have re-done
e.
Maria
e
je simpri completaminti sense
Mary
SCL is always completely
without
“Mary is always completely pennyless”
dut
everything
completaminti
completely
56
bês
money
f.
*Maria
Mary
e
je
SCL is
g.
Maria
no je plui
simpri cioche
Mary
neg is anymore always drunk
“Mary is not anymore always drunk at midday”
a
at
misdì
midday
h.
*Maria
Mary
a
at
misdì
midday
i.
No mangiave zà
plui
cjarn a Pasche
Neg ate
already anymore meat at Easter
“She did not eat meat anymore already at Easter”
j.
*No mangiave plui
zà
Neg ate
anymore already
cjarn a
meat at
k.
No an
mingul zà
Neg have
NegPol already
“They haven’t already phoned”
clamât
phoned
l.
*No an
Neg have
clamât
phoned
no je
neg is
completaminti simpri
completely
always
simpri
alway
zà
already
sense
without
plui
cioche
anymore drunk
mingul
NegPol
bês
money
Pasche
Easter
(62) a and b show that dut precedes bene; (62) c and d that completaminti
precedes dut; (62) e and f that simpri precedes completaminti; (62) g and h that
simpri precedes plui;(62) i and j that plui precedes zà; (62) k and l that mingul
precedes zà. Given the transitivity of sequentiality , the respective order of these
adverbials in Friulian is just the same as in SI, and is given in (63):
(63) mingul > zà > plui > simpri > completaminti > dut > ben
57
Now that this order has been confirmed for Friulian, we can check the positioning
of the PtPPL in the two languages and compare the results.
Basing his observations on empirical cross linguistic evidence, Cinque
(forthcoming) suggests that UG does not allow variation among languages with
respect to the type of functional projections they admit. He interprets the rigid
order of the above mentioned adverbials as a clear indication that they occupy fixed
specifier positions. This allows him to propose a hierarchy of adverbials - and,
consequently, of functional projections - that holds across languages. In his own
words:
“... adverbs are the overt manifestation of (the specifiers) of different functional
projections, which in certain languages may also manifest themselves via overt
material in the corresponding head positions”
(forthcoming: v).
Each of these heads finds in Cinque’s system a specific projection.
(64) - taken from Cinque (forthcoming: 77) - shows the positions where a PtPPL is
allowed to appear, represented by ticks, and those where it cannot do so,
represented by stars 26:
(64)
mica già più sempre completamente tutto * bene *
27
In Friulian, too, the PtPPL cannot be placed after tutto or bene:
(65) a.
E vin risistemât
dut
SCL have re-arranged everything
“We have re-arranged everything well”
ben
well
Northern Italian speakers find “odd” the position of the PtPPL when it
precedes mica and già; Southern speakers, on the other hand, prefer the preceding
position of the PtPPL to the one when it follows mica and già. This is really due to
regional variation and it does not interfere with the trigger of agreement, i.e. the
agreeing form of the PtPPL can be positioned either at the left or right of these two
adverbials.
26
The PtPPL can appear after bene and tutto only when an intonational break is
present before and after the relevant adverbial, which indicates extraposition of the
adverbial. The same observation holds true for Friulian.
27
58
b.
*E vin
SCL have
dut
everything
risistemât
re-arranged
c.
*E vin
SCL have
dut
everything
ben
well
ben
well
risistemât
re-arranged
In order to capture and define the difference between SI, younger Friulian speakers
(YF) and older Friulian speakers (OF) and understand the nature of the change in
progress in Friulian, representatives of two different Friulian generations were
tested 28. YF use the non-agreeing form of the PtPPL with a full pvDP, just like SI
does, whereas OF use the agreeing form. Comparing minimal pairs it was clear that
the position of the PtPPL and the appearance of the agreement morphology on it is
not correlated in Friulian, i.e. the relative position of the PtPPL with respect to the
adverbials does not vary depending on whether or not agreement with a full pvDP
is triggered. An example is given in (66). (66) a is from OF and shows agreement;
(66) b is from YF and it does not.
(66) a.
E vin
completaminti finude
le torte
OF
SCL have
completely
finished(f, sg) the cake (f, sg)
“We have completely finished the cake”
b.
E vin
SCL have
completaminti finût
le torte
completely
finished (UM) the cake
YF
This is a very significant observation as it suggests that PtPPL agr is not a result of
the PtPPL raising into an AGR functional projection. Some more support comes
from the PtPPL’s obligatory movement: (67) illustrates the positions where the
PtPPL is allowed to appear in Friulian; again, ticks stand for grammatical
sentences and stars for ungrammatical ones:
These informants spoke the same variety - geographically speaking - of
Friulian but belonged to two different generations, an older and a younger one.
28
59
(67)
mingul zà plui simpri completaminti dut * ben *
Do not forget that all these adverbials fill the Specifier position of functional heads,
so that ticks and stars indicate the head positions where the PtPPL is allowed or
banned, respectively. As in SI, in Friulian the PtPPL seems to have to raise to the
head to the left of dut, after having passed through the head to the right of ben and
that in between dut and ben. All other movements are optional.
Cinque (forthcoming: 173) speculates on this obligatory raising into what in his
system is an Aspectual functional projection hosting completaminti 29 and
interprets it as a necessary step that the PtPPL has to take in order to check its
“perfect” feature.
Belletti (1990: 13), Pollock (1989: 413), Chomsky (1989), Zanuttini (1991: 84) and
Giorgi & Pianesi (1991: 194) have all claimed, for independent reasons, that the
PtPPL lacks a functional projection containing [tense] features: the perfective
morpheme -t- (comparable to the English -ed morpheme) is seen as an expression
of aspectuality.
Following Baker’s (1985) “Mirror Principle”, Belletti (1990) assumes two functional
projections dominating the PtPPL: AgrP, carrying [gender, number] features and
AspP, containing the perfective morpheme. The AgrP dominates the AspP which,
in turn, selects the VP as its complement, as (69) illustrates (from Belletti (1990:
34)):
In order to give the reader an idea of the relative positions filled by all these
adverbs, I will include a partial version - only the relevant stretch of projections will
be included - of an approximate universal hierarchy of clausal functional
projections, formulated by Cinque (forthcoming: 178). The relevant adverbials are
in bold.
(68) ... [always Asp Perfect (?) [just Asp Retrospective [...
[completely Asp SgCompletive (I) [tutto AspPl Completive [well Voice [...
The question mark next to Asp Perfect indicates that it is unclear whether the
projection should be related to Asp Perfect or Asp Imperfect. This does not affect the
conclusions put forward here since that projection is higher than the one crucially
involved in obligatory PtPPL raising. I will gloss over the nature of the two types of
Completive Aspect projections since they are not directly relevant to my discussion.
Tutto is assumed by Cinque to be in a derived position, but this does not affect the
conclusions reached here.
29
60
(69)
Agr
AgrP
Agr’
[gender, number]
AspP
Asp
Asp’
-t-
VP
V
Adopting Cinque’s rigid order of functional projections 30 and assuming that the
aspectual feature is checked in the projection whose Specifier hosts completely,
and adopting also Belletti’s structure (69), the conclusion we reach is the following.
Under current theory, the function of the Agreement projection - be it AgrO or
AgrP - is to check the PtPPL’s agreement features. Its position - Cf (69) - must be
higher than Cinque’s Asp SgCompletive (I), and in order for PtPPL agr to be
instantiated, the PtPPL must raise into it. We have seen, on the other hand, that
the only obligatory movement the PtPPL undergoes is into Asp SgCompletive (I), and
crucially the PtPPL does not need to raise any further to show agreement, neither
in SI nor in Friulian, as shown by examples (70) a - DO cliticisation in SI-, b PtPPL agr with a full pvDP in Friulian - and c - DO cliticisation in Friulian.
Therefore there is no Agreement functional projection purely devoted to the
checking of agreement features.
(70) a.
b.
Le
abbiamo completamente mangiate
Them (f, pl) have
completely
eaten (f, pl)
“We have completely eaten them”
E vin completaminti
mangiadis
SCL have completely
eaten (f, pl)
“We have completely eaten the courgettes”
t
i
li cussutis
the courgettes (f, pl)
I will interpret the rich hierarchy of the full version of the structure
represented in footnote 29 to be an exhaustive and comprehensive array of all the
functional categories existing in a clause.
30
61
c.
Li
vin completaminti mangiadis
Them (f, pl) have completely
eaten (f, pl)
“We have completely eaten them”
i
t
i
Let us recapitulate on the evidence we have found and on what bearings it has on
an analysis of PtPPL agr.
First of all comparing SI and Friulian data we have seen that they both share the
same order in a given sequence of adverbials and that both require that the PtPPL
rise obligatorily to the head of the projection whose specifier hosts the adverbial
completely. We have then adopted Cinque’s hierarchical structure for adverbials
and his speculation that by rising into Asp SgCompletive (I) the PtPPL checks its
aspectual features. Were an Agreement projection present, it would be higher than
Asp SgCompletive (I) and PtPPL agr would crucially be dependent on the PtPPL
raising into it. The examples in (70) show that this is empirically falsified.
The conclusion that we can draw from this is that PtPPL agr is NOT triggered by
virtue of the PtPPL rising into an Agreement functional projection. By taking this
strong position, I want to propose an analysis of PtPPL agr where the agreement
instantiated between two elements is determined by a mechanism of feature
copying. In particular, agreement - as far as the one specified for [gender, number]
in concerned 31 - is a relational property.
Another consequence of the empirical evidence we have collected is that the
agreement morphology on the PtPPL is derived, in SI, by a [Spec, head]
configuration between the PtPPL and its cliticised internal argument in Asp
SgCompletive (I).
The next section presents the analysis I would like to put forward to account for
PtPPL agreement. The mechanism involved will be extended to DP internal
agreement, thus capturing their semantic affinity.
A tentative account of agreement specified for [person, number] features will
be given in section 3.5.
31
62
3.2 -
Theoretical assumptions
The concept of agreement as an instantiation of feature copying is a traditional
notion and has been widely exploited in the generative framework. The
assumptions underlying this idea are basically two: (i) the nominal source of
agreement is fully specified with regard to agreement features, and (ii) the
agreement relation ensures that duplicate (identical) information occurs on the
agreement target.
The notion of feature copying expressed in these terms covers those analyses
formulated in the Government-Binding (GB) framework, based on binding
relations (Chomsky (1981), Bouchard (1984)), although feature copying has been
replaced by a co-indexing relation between the nominal source and AGR or INFL.
Co-indexing and feature copying are equivalent in that they both ensure identity of
agreement features, as expressed in Chomsky (1981: 211):
“... when a NP and a pronominal are co-indexed, they must share the appropriate
features...”
Little is said in GB theories about types of agreement other than subject-verb
agreement, and a clear analysis of DP internal agreement has not been provided
yet.
One of the earliest accounts of DP internal agreement making use of the featurecopying theory goes back to the early transformational period and was proposed by
Postal (1964). Postal accounts for DP internal agreement in Spanish (1964: 44-48)
where [gender, number] features are shared by determiner, adjective and noun
within the same DP by formulating a set of transformational rules. These ensure
that the relevant features are copied from the noun and adjoined onto the
determiner and the adjective.
In formulating my account of PtPPL agr I have adopted Postal’s structuralist
approach in trying to define the formal conditions underlying the phenomenon
rather than engaging myself in an investigation of the function of agreement in
natural language (Cf Lehmann(1982, 1988)). However, to his transformational
63
rules I have substituted a domain in which feature copying can take place
(Grimshaw (1991).
On a theoretical level, my proposal exploits the notions of “extended projection”
formulated in a GB framework, and that of feature movement, promoted by
Chomsky’s (1995) “Minimalist Program”.
Before turning to the account of PtPPL agr and adjectival agreement that I would
like to suggest, I will illustrate the fundamental points on which my proposal rests.
These are briefly summarised in (71):
(71) a. PtPPL agr is not triggered by movement into a functional category;
b. The PtPPL is specified for [+V,+N] features in Friulian and for [+V, ~N]
in SI, which means that the [N] feature is present in SI but has no
value32;
c. in Friulian, DP internal agreement shares with PtPPL the same
mechanical process of instantiation;
d. [+N] features in languages where nouns are specified for [gender] must
be matched within the proper domain.
(71) a is the conclusion reached after having analysed the facts for PtPPL agr
illustrated in section 3.1.
(71) b is a hypothesis at this stage and will find some support in chapter 4 where a
diachronic account of the development undergone by the PtPPL and the auxiliary
“have” during the evolution from Classical Latin to Romance Languages is
presented. It must be clarified that by assuming a [+V,+N] feature specification for
the Friulian PtPPLs I am by no means claiming that these latter are identical to
adjectives. In fact the affinity that I claim exists between them is captured in (71) c:
they both share the same mechanism for the instantiation of agreement. On the
other hand, their different behaviour - outlined in section 3.4 - is captured in the
different definition of the domain in which this mechanism is operational.
Finally, (71) d is the underlying rule which triggers the agreement. [+N] features
must be shared by all the elements specified for them and contained in a “proper
domain”. The notion of “proper domain” must be defined with precision and
This kind of specification was originally suggested by Chomsky (1970), where
all lexical categories are specified with respect to [V] and [N] features.
32
64
accuracy, and in order to achieve this aim I will refer to Grimshaw’s (1991) notion
of “extended projection”.
I will follow Grimshaw in assuming that a functional projection D and its
complement NP form an extended projection. The underlying assumption is that D
and N share the same categorial features, [-V,+N] 33, and that the functional
element dominates the lexical one. The “Extended Projection” (1991: 4) and
the“Generalised Θ-Criterion” (1991: 9) rule out ungrammatical combinations.
Grimshaw formulates the “Extended Projection” as follows:
(72)
i.
ii.
iii.
iv.
x is the extended head of y, and y is an extended projection of x iff:
y dominates x,
y and x share all categorial features,
all nodes intervening between x and y share all categorial features,
if x and y are not in the same projection, the F value of y is higher than
the F value of x 34.
Let us apply this to the projection formed by the DP and the NP. D is a functional
element dominating the NP, both D and N are specified for the same categorial
features [-V,+N], and D’s {F} value is higher than N’s, so that the conditions are
satisfied. I will here take “intervening” in clause iii to refer to an intervening
projection. (73) illustrates the structure of the extended projection of N, with an AP
adjoined to the NP, and a PP complement to the NP. The relevant nodes that are
immediately contained in the extended projection are encircled.
Grimshaw extends the feature system formulated by Chomsky (1970) to
functional categories, so that N and D are [-V,+N] and V and I are [+V,-N]. Also, I
follow Abney (1987), Szabolcsi (1987), Fassi Fehri (1989) among others in adopting
the DP hypothesis.
33
“F” represents the functional feature associated to the element. Grimshaw
assigns the value {F0} to a lexical category and {F1} to a functional one, so that by
iv. only the configuration where the functional category dominated the lexical one
could become an extended projection.
34
65
(73)
We will see how this definition can account for DP internal agreement in the next
section.
3.3 -
The structure
As we have seen in section 1.5.1, in both SI and Friulian within a DP a noun, an
adjective and a determiner all agree in [gender, number]. When the NP has a PP
complement, as in “La bella fotografia del gatto” - the beautiful photograph of the
cat - the PP complement does not share the same features as the rest of the PP (the
photograph is feminine, as are the D la and the AP bella, while the cat is
masculine). Similarly, as shown in examples (74) a and b, c and d from SI and
Friulian respectively, an AdvP modifying the AP does not share the same features
as the rest of the DP:
(74) a.
b.
La
casa
molto
The (f, sg) house (f, sg) very
“The very ancient house”
*La
casa
The (f, sg) house (f, sg)
antica
ancient (f, sg)
molta
antica
very (f, sg) ancient (f, sg)
66
c.
Le
cjase
The (f, sg) house (f, sg)
tant
very
antiche
ancient (f, sg)
d.
*Le
cjase
The (f, sg) house (f, sg)
tante
antiche
very (f, sg) ancient (f, sg)
Since we are taking DP internal agreement to be a process of feature copying, the
algorithm setting the conditions for the instantiation of this process must allow the
D and the AP to be targets of the feature-copying process, and it must exclude
AdvPs and PPs complements to the NP. (75) expresses this set of requirements
which will yield the desired result:
(75) Copy the features [ψ] from x to y iff:
i. x and y are inherently specified for [+N], and
ii. y is immediately dominated by a node immediately contained in x’s
extended projection.
Comparing (73) and (75), we see that the nodes which are immediately contained
in N’s extended projection are DP, D’, NP, N’ and N. If W is immediately dominated
by a node Z it means that Z is the first node dominating W. According to this, D
and AP are legitimate targets for the copying-feature process being immediately
dominated by, respectively, D’ and NP, which are both immediately contained in
the extended projection. AdvPs, on the other hand, not being immediately
dominated by a node immediately contained in the extended projection, are
excluded. Turning now to the PP complement to the NP, we notice that being
immediately dominated by N’ the PP is a potential target for the agreement, but
being categorically specified for [-N] it does not satisfy condition (75) i. Were a
language to allow for DP complements to the NP, our algorithm predicts that they
should be targets for the agreement. On similar lines, [Spec, DP] is a potential
target, too, allowing for whatever elements appearing there and inherently
specified for [+N] to share the features with the rest of the DP 35.
In (73) for
This prediction allows us to account for DP internal agreement where APs are
placed in [Spec, DP], Cf Cinque (1994a).
35
67
simplicity I have glossed over any functional projections which have been assumed
to exist inside the DP 36. The movement of the N to functional projections will not
affect the agreement mechanism, since it is assumed that Agr is instantiated
irrespective of N-movement, and can , therefore, take place before or after it. Also,
whether the AP is adjoined to the left or to the right of the NP does not affect the
agreement process just illustrated, given that it will be always immediately
dominated by the NP.
In summary, we have seen that the notions of extended projection, featurecopying, inherent specification for [gender, number] together with the algorithm in
(75) can satisfactorily account for DP internal agreement. (76) recapitulates: the
nodes marked by a star are the potential targets for the feature-copying process.
Whether they actually are targeted depends on their specification.
(76)
The reader is referred to Cinque (1994a) where the author argues for partial
evidence of N-movement within the Romance DP, discarding evidence brought
forward by Lamarche (1991). Cf also Ritter (1991, 1993) who argues, respectively,
for two functional projections within the Romance DP and that [gender, number]
features are placed in NumP, a functional projection dominating NP; Picallo (1991)
who proposes that N raises into NumP in Catalan; Bianchi & Figueiredo-Silva
(1994) argue for the existence of two functional projections above the NP, NumP
and GenP respectively, in Brazilian Portuguese and SI. See also Bernstein (1991) for
the idea that the N raises to different positions in different languages and that this
difference is parametrised.
36
68
Let us turn now to PtPPL agr.
As it has already been claimed in (71), PtPPL agr shares with DP internal
agreement the mechanism which instantiates it, i.e. feature copying from the
nominal source on the target element. In PtPPL agr the target element is, of course,
the PtPPL. In Friulian the instantiation of PtPPL agr crucially depends on the
feature specification carried by the PtPPL, which is different from the one carried
by SI PtPPL. This difference is responsible for the lack of PtPPL agr in SI with a full
pvDP.
(77) is the structure I adopt for a transitive PtPPL:
(77)
...
Spec
AspP
Asp’
Asp
VP
Spec
V’
V
DP
Spec
D’
D
NP
[+N] [+N]
The PtPPL is placed in V, which will carry [+V,+N] features in Friulian, [+V,~N] in
SI. Under Grimshaw’s approach, a verb and its complement never form an
extended projection because they crucially do not share the same categorial
features and because the requirement that the functional projection govern the
lexical one is not met 37.
There is an alternative solution. If AspP and VP were taken to form an
extended projection - both being specified for the same categorial features, as
Grimshaw does for I and V -, then the verbs’ internal argument would be
immediately dominated by a node immediately contained in the extended
69
37
Let us consider the Friulian case first.
It is apparent that there is no movement involved in the instantiation of PtPPL agr
with a full pvDP. Therefore, I would like to suggest that PtPPL agr is triggered by
the PtPPL “attracting” a copy of the inherent [+N] features of the closest element it
c-commands.
The necessity for the requirement that the nominal source of agreement be ccommanded by the PtPPL, stems from the fact that the PtPPL never agrees with its
pre-verbal subject. The c-commanding restriction accounts for all instances of
PtPPL agr in unaccusative, ergative and passive constructions, where the logical
subject is base generated post-verbally, as well as in transitive constructions.
projection, as shown in (78). This, however, would not allow us to capture the
different behaviour of PtPPLs and APs, discussed in section 3.4.
(78):
70
In SI, on the other hand, the PtPPL is categorically specified as [+V, ~N], i.e. the
feature [N] is present but has no value 38. This lack of value is reflected in the fact
that the PtPPL does not attract a copy of its DP complement’s [gender, number]
specification. Once positioned in a [Spec, head] configuration with its internal
argument, though, the PtPPL agrees with it, since the two elements carry
compatible features, [+N] and [~N]. This accounts for all instances of PtPPL agr in
SI, i.e. cliticised, unaccusative, ergative and passive constructions.
The PtPPL and its internal argument will be in a [Spec, head] configuration once
raised in the functional projection dominating the PtPPL, AspP or, in Cinque’s
(forthcoming) terms, Asp SgCompletive (I), and the PtPPL’s feature [~N] will be
turned into [+N].
(79) and (80) represents PtPPL agr in, respectively, Friulian and SI. The features
carried by the V’s trace in (80) show that the [N] feature on the PTPPL changes
from having no value to being specified for the same value as the DP’s with which it
is in a [Spec, head] configuration.
(79)
...
Spec
AspP
Asp’
Asp
VP
Spec
V
V’
[+V,+N]
PtPPL
DP
[-V,+N]
[gender, number] are parasitic on the feature [N], i.e. in order to have
inherent [gender, number] features on a lexical item, that lexical item has to be
[+N]. These same features can then be copied onto other lexical items.
With the specification [+V, ~N] the PtPPL is distinguished from all other verbal
forms which are [+V,-N].
38
71
(80)
...
Spec
DP
[-V,+N]
i
AspP
Asp’
Asp
PtPPLj
[+V,+N]
VP
Spec
V
V’
[+V, ~N] t
j
DP
[-V,+N] t
i
(81) is the algorithm that expresses formally the process of PtPPL agr:
(81)
Copy the features [ψ] from x to y iff:
either
i. x and y are inherently specified for [+N], and
ii. x is the closest node c-commanded by y,
or
iii. x and y have compatible feature specification, and
iv. they are in a [Spec, head] configuration.
The first set of conditions accounts for PtPPL agr in Friulian and the second for all
the other Romance languages. It is crucial to underline that “compatible” features
here is taken to mean that the value of the two features does not clash. Thus two
elements will have compatible features if one carries a positive value for that
feature and the other carries no value at all. On the other hand they will not be
compatible if one carries a negative value and the other a positive one. As a
consequence, this condition predicts that if the PtPPL is specified as [+V, -N] in a
language, then no agreement will be possible. This is indeed what happens in
Rumanian, Spanish and Portuguese (Cf chapter 4).
Finally, it must be pointed out that in accounting for PtPPL agr triggered by
cliticisation in SI I will not follow Chomsky’s (1988) adaptation of Kayne’s (1985)
original proposal that the clitic is adjoined to AgrO and creates a Spec-like position
which triggers agreement with AgrO. Rather, I will follow Sportiche (1996) in
assuming that the whole DP rises into [Spec, Asp] and agreement is triggered with
72
the PtPPL raising into the head position 39. On the last step of its movement, the D
clitic will be extracted from the DP and head will move to reach its final
destination40.
In summary, this section has presented a unified account for DP internal and
PtPPL agr in Friulian. The unifying thread between the two is the necessity for
items carrying [+N] features within the proper domain to share them. The
distinguishing factor between the two, is the definition of proper domain: this
latter is delimited in the DP to the same projection, in the case of PtPPL agr to the
attraction of a copy of the [+N] features from the closest c-commanded head.
In the next section the importance of a different domain proves important when
accounting for the differences in behaviour between APs and PtPPLs agr with lists.
With respect to raising into the specifier of a functional category, Cinque’s
(forthcoming) proposal is problematic since all specifier positions are filled by
adverbials. In order to allow for the raising of the DP we must assume that either
there are more projections with empty specifiers - which would undermine the
value of the observations in section 3.1, or allow for multiple specifiers. This issue
will not be pursued here any further.
39
Kayne (1989: part I, 1991: sect. 1.1) convincingly shows that clitics adjoin to
functional heads.
40
73
3.4 -
Relevance of the different domains of the featurecopying process in PtPPL and DP internal agr
It has already been shown in section 1.5.1 (Cf examples (29) a and b) that Friulian
PtPPLs, just as SI ones, assign Case and a Θ-role to their internal argument41, and a
Θ-role to their external argument. Adjectives do not assign Case or Θ-roles 42 .
Generally speaking, PtPPLs used adjectivally inside the DP cannot be modified by
intensifiers, while APs can. Some examples from Friulian 43:
Following Giorgi & Pianesi (1991) I will assume that since an auxiliary does
not take part in the semantic meaning of the composed tense but only in the
syntactic one, it is base generated under T. Thus it wouldn’t be able to assign an
external Θ-role on its own, i.e. it would be unspecified for external Θ-role
assignment (in Lois (1990)’s sense). The actor of the action would be Θ-marked by
the PtPPL. The auxiliary here is thought to be only “helping” the PtPPL to assign a
Θ-role to the subject. As noted in Stowell (1982) the auxiliary and the PtPPL share
the same agent, so that it is reasonable to believe the Θ-role is assigned
compositionally.
41
Adjectives can, nevertheless, express the external Θ-role of a N. (Cf Giorgi
and Longobardi (1991: 125ff) and Kayne (1981: 111)).
42
43
The reader is referred to Bisetto (1994) for a set of criteria to distinguish
74
(82) a.
b.
*I tant
The very
I
tant
The very
lavâs
washed (PtPPL)
afetuôs
affectionate (AP)
barcôns
windows
frus
children
Furthermore, a difference in behaviour is found when lists of items are involved.
In SI, and in Friulian too, in the presence of a mixed group of people belonging to
both sexes an adjective referring to it must carry a [m, pl] ending, as shown by (83)
a from SI and (83) b from Friulian:
(83) a.
Che bei
ragazzi
e
ragazze in
What beautiful [m, pl] boys [m, pl] and girls [f, pl] in
“How beautiful boys and girls in traditional costumes!”
b.
Se
What
biei
frus e frutis
beautiful boys and girls
in
in
costume!
costume
costum!
costume
When the PtPPL refers to a group of items carrying different grammatical gender,
the situation is different: in both languages the PtPPL surfaces in its unmarked [m,
sg] form, and the sentence results in ungrammaticality if we apply to PtPPLs the
between “verbal” and “adjectival” PtPPLs in SI. Looking at the semantics of the
verbs, Bisetto distinguishes between those verbs which will cause a “resultative
state” in the object - examples are to break, to wash, to redden, etc. - and those
which do not - to reach, to transport, to shout. The former have both a PtPPL and
an adjective, the latter only a PtPPL. The two forms, adjectival and verbal, are the
result of two different processes of word formation, but have suffixes which are
phonologically identical.
75
same rule adjectives follow, i.e. in a group of mixed gender or sexes use the [m, pl]
ending. This is shown by (84) for SI and (85) for Friulian:
(84) a. Ho visto/
*visti
Marco, Luisa,
Giorgio e
Ada
Have seen (UM)/*(m, pl) Marco, Luisa,
Giorgio and Ada
“I have seen Marco, Luisa, Giorgio and Ada”
b.
Ho comprato/ *comprati i fichi,
le pesche,
Have bought (UM)/*(m, pl)
the figs (m, pl), the peaches (f, pl),
i pompelmi,
le pere
e
i meloni
the grapefruits (m, pl), the pears (f, pl) and the melons (m, pl)
“I have bought figs, peaches, grapefruits, pears and melons”
(85) a.
b.
E ai viudût/ *viudûs Marco, Luisa, Giorgio e Ada
SCL have seen (UM)/*(m, pl) Marco, Luisa, Giorgio and Ada
“I have seen Marco, Luisa, Giorgio and Ada”
E ai comprât/
*comprâs i fics,
li pomis,
SCL have bought (UM)/*(m, pl)
figs (m, pl),
the apples (f, pl),
i pompelmos,
li fragulis
e i melôns
grapefruits (m, pl), the strawberries (f, pl) and the melons (m, pl)
“I have bought figs, apples, grapefruits, strawberries and melons”
In examples a in (84) and (85) each DP in the list has an empty D; in examples b in
(84) and (85) each DP in the list has its own D, and it is different from the DP
which follows and precedes it. This shows that the agreement is not affected by
whether Ds are present or not in the lists.
Under the theory formulated in this chapter this difference can be captured by the
different feature-copying operational domains assumed for DP internal and PtPPL
agr.
It appears that while adjectival agreement between a NP and an AP can “see”
[number] features when dealing with lists, PtPPL cannot do so when their internal
76
argument is a list. I would like to suggest that this depends on the AP and the NP
belonging to the same extended projection and on the necessity for them to be
specified for the same ψ features. As far as [gender] is concerned, both in SI and
Friulian, a group of items of mixed sexes or genders is referred to as [m] masculine being the unmarked choice for gender. The resulting specification which
appears on the adjective, [m, pl], is derived as follows: the specification for
[gender] is the default unmarked one, while the [number] specification comes from
the group containing a number of individuals, and being, therefore, plural.
Belonging to the same extended projection is the crucial factor here, since it forces
plurality on the A and D.
PtPPL agr, on the other hand, operates across projections on its selected internal
argument. When the PtPPL selects for a list formed by many DPs as its internal
argument, each of these DPs has its own feature specification, so that the feature
specification cannot be represented by one set of features. Furthermore, all these
DPs are all c-commanded by the PtPPL so that they are all on an equal hierarchical
position with respect to the PtPPL, i.e. the DPs are all sisters to the PtPPL 44. As a
consequence, the PtPPL is not able to “choose” which copy of features to attract,
and it surfaces in its [m, sg] unmarked form.
Although suggestive rather than conclusive, this analysis shows how the different
operational domain assumed for PtPPL and DP internal agreement could capture
the above presented different behaviour of PtPPLs and APs.
The last section in this chapter will present some speculations on the nature of
agreement specified for [person] features and draw a conclusion for the chapter.
3.5 -
Final remarks and Conclusion
The semantic affinity between PtPPLs and adjectives has been interpreted in
44
I will assume that the syntactic representation of a list is a recursive DP.
77
syntactic terms in having the two categories sharing the same mechanism for the
instantiation of agreement with a nominal. In the previous section, the operational
domain of this mechanism has been proved to be very significant when dealing
with agreement in special circumstances, such as with lists.
The investigation of PtPPL agr was triggered by the original observation of the
different feature specification of the two functional nodes allegedly involved in
agreement - AgrS and AgrO. In this chapter some convincing evidence against the
existence of AgrO has been presented. Given that AgrO is specified for [gender,
number], at this stage it is desirable to make a further suggestion, which takes
shape in two possible claims: either the mechanism involved in agreement for
[gender] features is different to the one involved in [person] features, due to the
existence in the second case of AgrS, or agreement is only a relation in ALL
circumstances, and there exist no AGRPs. It must be stressed that these are only
speculations and should be tested cross linguistically to see whether they hold.
The first possibility sees verbal agreement - V agr - as a property acquired by the
verb by virtue of raising into an AgrS head position. This claim would be supported
by some evidence from the Scandinavian family, where the existence of agreement
in [person] features within the verbal paradigm implies the existence of an AgrS
node to which V raises (cf Holmberg & Platzack (1995)).
Alternatively, we could follow Iatridou (1990) is assuming that there is no such
functional projection as AgrS and that the function of Case assigning - or checking
- is performed by Tense. This is supported by the fact that in Friulian and SI only a
finite verb can assign Nominative Case to its logical subject and from the fact that a
finite verb shows inflection for [person] features, we could deduce that [person]
features are involved in Case assigning processes. Thus the sharing of the features
between a finite verb and its syntactic subject would surface as a result of a [Spec,
head] configuration between the two. A consequence of this is that by eliminating
all the Agreement functional projection the tension between functional projections
whose semantic properties are clear - such as Tense or Negation - and Agreement
functional projections, which do not have a property that can be isolated
semantically, could be resolved. By following this second claim, agreement can
therefore be reduced to a pure relation between two or more elements which
78
determines feature sharing between them.
In summary, this chapter has investigated the relative positions of PtPPL, its object
and some adverbials, in order to establish whether PtPPL agr is triggered as a
result of the PtPPL raising into an Agreement functional projection. This was
discarded due to the evidence that PtPPL moves obligatorily into one functional
projection, AspP, but no further movement is necessary in order to trigger
agreement. In order to capture the semantic affinity between PtPPLs and
adjectives, it has been proposed that the same mechanism which is responsible for
PtPPL agr is also responsible for adjectival agreement.
A possible rule has been formulated with precise restrictions imposed by
configurational relations, invoking the concepts extended projection, immediately
contained and immediately dominated for adjectival agreement, and c-command
for PtPPL agr. The matching of the [+N] features underlies both agreements, and
the difference between SI and Friulian PtPPL agr has been captured in claiming
that the two verbal forms carry different feature specifications in the two
languages. While PtPPL in SI is a [+V, ~N] category, it is a [+V,+N] category in
Friulian, due to the fact that Friulian is still undergoing the process which derived
the modern PtPPLs from Latin.
The system so formulated has been then applied to a problematic case, agreement
with lists, and a solution exploiting the two different operational domains has been
proposed. Finally, some speculations on the difference between agreement in
[person] and agreement in [gender] features have been made, suggesting two
alternatives: either agreement in [person] features reflects the existence of an AgrS
node while agreement in [gender] is only a relation, or simply all agreement is a
relation. If this last suggestion finds some support from a cross-linguistic
investigation, then some more evidence can be available to decide on the long
debated issue of whether agreement is a relation or a functional projection.
The next chapter will present some support for the claim made on the difference
categorial feature specification of the PtPPL in SI and Friulian.
79
Chapter 4
Historical development of the PtPPL and
the auxiliary in Romance Languages
4.0
4.1
4.2
4.3
4.4
4.5
Introduction
The Latin tense system
How did the change come about?
Effects on word order: evidence from Portuguese and Friulian
The status of Friulian with respect to the other Romance Languages
Conclusion
4.0 -
Introduction
In the previous chapter it has been claimed that the different behaviour of the
PtPPL in SI and in Friulian is due to the different feature specification the PtPPL
has in the two languages. While in SI it is specified for [+V, ~N] features, it is
specified for [+V, +N] in Friulian. The fact that Friulian is in a way more
conservative than SI in allowing structures considered archaic in Modern Italian,
suggests that Friulian is still undergoing the diachronic change already completed
in SI.
Taking this observation as a starting point in this chapter we will find some
more support for this claim by investigating what has been considered to be the
"ancestor" of the compound tenses in Latin.
Section 4.1 presents some data from Latin illustrating its verbal tense system,
80
how analytic forms developed, what function they had and the semantic change
they underwent. Section 4.2 investigates the syntactic reflection of the semantic
changes. Section 4.3 focuses on the changes undergone by habere - to have - and
the influence this transformation had on word order. Some data from Friulian and
Portuguese shows a close link between the function performed by the respective
version of habere, linear order of auxiliary - PtPPL - DO and the possibility of the
subject of habere and of the PtPPL of not being co-referential. Section 4.4
interprets the facts illustrated in section 4.3 and places Friulian in the wider
framework of Romance Languages.
4.1 -
The Latin tense system
The Classical Latin tense system had the means of expressing both the
location of an event in present, past or future and also its relation to a specific
reference point, i.e. anterior, simultaneous or posterior to it. The former function
was performed by the so-called synthetic one-word forms, such as scripsi - I wrote,
lexi - I read. The latter was performed by analytic, or periphrastic forms, found in
the passive construction or with verbs called deponent, i.e. which had only passive
morphology. These were fully operational throughout Latin. Cf (86):
(86) a.
laudatus
eram
praised
I was
"I had been praised"
b.
profectus eram
left
I was
"I had left"
In analytic constructions only one auxiliary was used, esse - to be, a one-place
predicate which gradually lost its original existential meaning and came to
represent only a general relation. In this respect it can be claimed that esse was a
real auxiliary, resembling its French and SI modern versions. The PtPPL was
passive in meaning (deponents excepted) and was “... a verbal adjective [which
agreed] in gender, number and Case with the N it modified...” (Randall (1992:
233)).
81
Some constructions with habere - to have - and the PtPPL were present in
Latin, too, and are considered to be the "ancestors" of the Modern Romance
constructions, although they were not equivalent to them.
The Latin construction was formed by a finite form of habere followed by an
object and the passive PtPPL, as shown in example (87):
(87) qui eum vinctum
habebit
who him tied
will have
"Who shall hold him in bonds" (From Pinkster (1987: 197))
Some more examples are shown in (88): (88) c shows the θ-role and Case
assignment process:
(88) a.
in
ea
provincia pecunias
magnas
in
that
province
capital (f, pl)
great (f, pl)
collocatas
habent
invested (f, pl)
they have
“They have great capitals invested in that province”
(Cicero)
b.
habeo
epistulam scriptam
I-have
letter-ACC written-ACC
“I possess the written letter”
c.
Case
Case
|_________| |__________|
[ego]
habeo
epistulam
|_θ-role__|
|___θ-role_|
scriptam
in the above examples cannot be considered a proper auxiliary yet,
since it retains its meaning as a full possession verb and an argument structure of
its own, as can be seen from its ability of assigning Case and a θ-role to its subject
and object. The object is selected by habere and the PtPPL is really a
Habere
82
Praedicativum, i.e. it indicates a property in which the direct object is as a result of
the action predicated by the PtPPL, which, as pointed out above, syntactically
behaves like an attribute, agreeing in case, number and gender with the object. It is
interesting to notice that in Classical Latin only those PtPPLs which could act as
adjectives and mainly derived from resultative or terminative verbs were allowed in
these periphrastic constructions. In sentences like (87) and (88) a and b the PtPPL
could be omitted without affecting the general meaning of the sentence, i.e. (87)
can simply mean “who shall hold him”, (88) a “they have great capitals in that
province”, and (88) b “I possess the letter”.
Expressing this graphically, analytic constructions could be represented as:
(89) Habere
+
(Object + PtPPL)
On a parallel with such constructions, there existed very similar ones where
the PtPPL could not be omitted without making the sentence meaningless. An
example is given in (90):
(90) (Flamines) ... caput
cinctum
habebant filo
Flamines
head (m, sg) girted (m, sg) they had
by a string
"The Flamines had their heads girted by a woollen fillet" (From Pinkster
(1987: 1997)).
Since the omission of cinctum results in the sentence being pointlessly
obvious as not deserving to be overtly stated - “they had heads” -, it has been
claimed (Ramat (1987: 369)) that in these examples habere should be considered
as a three place predicate, where the PtPPL is one of them and predicates of the
state in which the object is as a consequence of the action. Graphically, these
constructions can be all reduced to the structure represented in (91):
(91) (Habere
+
participle)
+
object
Salvi (1987: 226) compares these structures to others which are found in SI.
83
The PtPPL performs here the same function as it did in Latin sentences (90). Some
examples, reported from Salvi (1987: 226) are given in (92):
(92) Tengo gli
occhi
aperti
I-keep the eyes (m, pl) open (m, pl)
“I keep the eyes open”
Some syntactic tests show that indeed gli occhi and aperti are two separate
constituents:
(92) a.
Li
tengo aperti
Them I-keep open
“I keep them open”
b.
Come tieni gli
occhi?
Aperti
How keep the eyes?
Open
“How do you keep your eyes? Open”
c.
Gli occhi che tengo aperti
The eyes that I keep open
"The eyes that I keep open"
In the Latin examples the object and the PtPPL are two separate constituents,
just like their SI counterparts, as can be seen in (93), similar to (92) c, reported in
Salvi (1987: 227):
(93)(equitatum) quem
ex
omni provincia... coactum
habebat
cavalry
whichACC from each province
gatheredACC
he had
"The cavalry which he had gathered from all
provinces"
From Cicero onwards, an increasing variety of verbs were allowed to enter into
the habere + PtPPL construction, perception and cognition verbs in particular. The
object could no more be understood as the object of habere, nor the PtPPL as
84
predicative of its change of status as a result of the action. In fact with cognition
and perception verbs it is the subject of the action which is undergoing a change,
i.e. it becomes aware of the existence of the object - as it has been pointed out in
chapter 2. Some examples are given in (94), taken from Pinkster (1987: 204):
(94)a.
b.
cum cognitum habeas
quod sit sumni
rectoris...
when known
you have
what be of supreme governor
numen
will
"When you have realised what is the will of the supreme lord"
auditum
habemus
heard
we have
"We have heard that ..."
quod ...
that ...
It was only in a later period that the periphrastic construction was extended to
intransitive verbs, so that habeo dormitum became to signify what it means
nowadays in SI, ho dormito - I have slept.
It is in these last types of constructions that the amalgamation of habere and
the PtPPL into a complex verb form was completed, giving origin to the Modern
Romance constructions. It is clear from the examples in (94) that the sentential
object is selected by auditum and not by habere, which, semantically bleached, has
come to assume the meaning of a general relation, similar to the one instantiated
by esse.
4.2 -
How did the change come about?
How did this transformation take place?
It began on a semantic level first. The original meaning of the Latin
periphrasis took on the meaning it has nowadays in Romance.
This was due to the semantic emptying of habere. From having the same
meaning as tenere - to keep, a durative action - it acquired the meaning of
possession, and finally it came to represent just a generic relation, i.e. to perform
the function of an auxiliary, with no selectional properties of its own, very similarly
85
to esse. If esse expressed a relation between two entities, A and B, habere could
express the reverse relation, between B and A. An example is found in pairs such as
:
(95)a.
b.
Domus
est Marco
House is
to Marcus
"The house is of Marcus"
Marcus
habet domum
Marcus
has house
"Marcus has a house" (adapted from Salvi (1987: 229))
Another factor that contributed to the change was the increasing tendency of
the subjects of the two verbal forms to coincide, particularly evident in
constructions containing perception and cognitive verbs (Cf examples (94) a and b
above). Habere did not enter into the semantic interpretation of the action and this
caused the transfer of the semantic axis of the sentence from the auxiliary and its
object to the PtPPL, i.e. it became transparent to verbal restrictions. The
consequence of this is that the PtPPL came to lose its adjectival nature and assume
a verbal one. The object was then selected by the PtPPL, and the construction
which originally meant "I own the result of an action" became to mean "I
performed that action in the past".
The change was only later followed by a syntactic change. A first consequence
in interpretation was that the restrictions on the categories of verbs which were
allowed to enter into the construction were gradually lifted, so that resultative or
terminative verbs were sidedused alongside of perception, cognitive and activity
verbs. Thus the adjectival nature of the PtPPL gradually disappeared, and it was
ultimately reflected by the loss of the agreement between the PtPPL and a full
pvDP, which is what we find in SI.
From a structural point of view, it has been claimed (Salvi (1987: 228)) that
the syntactic structure the periphrasis
habeo
epistulam scriptam
86
had in Latin was a small clause governed by habere, where a predication relation
between the PtPPL and the object was instantiated. The small clause was headed
by the PtPPL functioning as an adjective. The object was assigned Accusative Case
by habere through the AP small clause, which did not represent a barrier.
The structure assigned to it is represented in (96) (Salvi (1987: 228)):
(96)
Spec
[ego]
VP
V
habeo
V'
AP
NP
A’
A
epistulam
scriptam
(96) shows that there is no direct connection between habeo and scriptam.
When the semantic emptying of habere was completed and the syntactic
transformation started to take place, the small clause was still headed by the
PtPPL, but its category changed: from being an AP the small clause turned into a
VP, which reflected the shift of the axis of the sentence on the PtPPL. The new
structure is represented in(97):
VP1
(97)
Spec
[ego]
V’
V1
habeo
VP2
NP
epistulam
V’
V2
scriptam
The subject of the small clause - epistulam - was now the object of the PtPPL,
so that adopting structure (98) instead of (97), a parallel can be drawn with passive
sentences . In (98) [Spec, VP2] is empty for the NP to raise into in order to get Acc
87
Case from habere - the PtPPL cannot yet do so.
VP1
(98)
Spec
[ego]
V’
V1
VP2
NP
habeo
V’
V2
NP
scriptam
epistulam
Structure (98) is based on the assumption that the PtPPL is not able to assign
an external Θ-role on its own. Habere, having undergone a semantic bleaching, can
no longer assign an external Θ-role on its own either, but can help the PtPPL to do
so.
Finally, when habere had completely lost its properties as a full verb, it could
no longer assign Acc Case to the object: the PtPPL acquired that property,
becoming a verbal form. As a consequence agreement with the object - which made
it so adjective-like - disappeared. Structure (99) represents the final development
of the Latin construction, which is nowadays found in the Romance languages 45:
(99)
Spec
[io]
VP
V
ho
V’
NP
VP
V
scritto
V’
NP
la lettera
In the next section some facts relative to the rigidification of the linear order
exemplify the ability of the auxiliary to assign Acc Case to the DP object. The
examples are drawn from Friulian and Portuguese.
All the above given structures were formulated by Salvi in the Government
and Binding framework.
45
88
4.3 - Effects on word order: evidence from
Portuguese and Friulian
The shift from the construction in (88), where the object and the PtPPL were
in a relation very similar to the one instantiated between a noun and an adjective
modifying it, to those where the auxiliary and the PtPPL came to amalgamate into a
complex verb form was reflected also in the linear order that the auxiliary, the
PtPPL and the object had with respect to each other.
When habere imposed selectional restrictions on the object, its subject and
the subject of the PtPPL did not necessarily coincide. (88), for example, means that
“I possess the result of the fact that a letter has been written [by somebody]”, and
not “I have written the letter myself”.
In later examples, illustrated in the previous section by (93) and (94), habere
and the PtPPL tended to share the same subject, so that a direct relation between
the two arose. This affected word order, so much so that the sequence that became
more and more rigid to the extent of being the only one acceptable in expressing an
action performed in the past was Aux - PtPPL - DO.
The same is also true nowadays. Different word order leads to a different
interpretation; in particular, when the object immediately follows the A aux this
latter is interpreted as a main verb expressing possession.
This is witnessed in Portuguese, investigated by Lois (1990). In this language
the two different functions of “to have” as an auxiliary and as a main verb are not
morphologically distinct in two corresponding different verbal forms (Cf Spanish
haber and tener respectively). Portuguese ter performs both, but it behaves
differently, syntactically speaking, depending on the function it performs:
agreement of the PtPPL with the object signals the function of ter as a main verb,
lack of it signals its function as an auxiliary. Compare (from Lois (1990: 236 n 4)):
(100)a.
Tenho a
carta
escrita
89
I-have the letter (f, sg)
“I possess the written letter”
b.
written (f, sg)
Tenho escrito
a
carta
I-have written (UM) the letter (f, sg)
“I have written the letter”
Notably, the position of the DP is crucial to the function of ter: (100) a becomes
ungrammatical if agreement is not instantiated:
(101)
*Tenho
a
carta escrito
A similar phenomenon can be observed in one of the varieties of Friulian that
uses the “non-agreeing” form of the PtPPL with full pvDPs - what has been referred
to as “young Friulian” in chapter 3. Here, too, the position of the DP disambiguates
the interpretation of avere, i.e. as a main verb or as an auxiliary. On a suggestion
by Poletto (personal communication) minimal pairs with the presence of a
sentential adverb were tested to see if there was a difference in meaning between
sentences with different word order. The pairs ae given in (102): it can be seen that
the position of the DO plays a crucial role:
(102)a.
E
ai
fumât
simpri
PtPPL [UM] Adv
SCL have
smoked
always
“I have always smoked cigarettes”
b.
E
ai
c.
E
ai
li
DO
the
sigaretis
cigarettes
simpri
fumât
li sigaretis
Adv
PtPPL [UM] DO
SCL have
always smoked
the cigarettes
“I have always smoked cigarettes”
li sigaretis
DO
simpri
Adv
90
fumadis
PtPPL [+agr]
SCL have the cigarettes always smoked
“I possess cigarettes always (already) smoked”
d.
E
ai
simpri
li sigaretis
fumadis
Adv
DO
PtPPL [+agr]
SCL have always
the cigarettes smoked
“I possess always (already) smoked cigarettes”
(102) a and b unambiguously yield an interpretation where the Friulian
counterpart of avere - vè - is an auxiliary and the PtPPL does not agree, just like
in SI.(102) c and d cannot be accepted without agreement, and vè comes to
function as the main possession verb: the DP is its internal argument and the
PtPPL is modifying it, just like an adjective. Interestingly, (102) d is preferred when
an agent is added, as in (103) 46:
(103) E
ai
simpri li
sigaretis
fumadis
SCL have always the cigarettes smoked
“I always have my cigarettes smoked by your dad”
da to pari
from your dad
(103) resembles very closely English constructions “to have something done
by somebody”, constructions possible in Friulian, but not in SI. Compare (104)
from Friulian and (105)a and b from SI:
(104) E
ai
vude le
cjase rimodernade
da
cantine fin
SCL have had the house re-decorated
from cellar up to
sul solâr tre ains fa
on the attic three years ago
“I have had my house redecorated from top to bottom three years ago”
(105) a.
*Ho avuto la
casa rimodernata da
cima a
fondo
In SI the counterpart of sentence (102) c and d are marginal and they have the
same meaning as in Friulian. They do not allow, however, for the addition of an
agent. Compare also the ungrammaticality of example (20) a.
46
91
Have had the house re-decorated from top to
“I have had the house re-decorated from top to bottom”
(105) b.
Mi
hanno rimodernato la
To-me
have re-decorated the
“They have re-decorated my house”
bottom
casa
house
In English, too, the position of to have and the object decides which function
the PtPPL is playing: verb or adjective. Denison (1993: 341) reaches similar
conclusions: after having investigated the nature of have as a main verb and as an
auxiliary he shows that adjacency of finite have and PtPPL gives the auxiliary
interpretation, while non-adjacency of the two leads to the main verb
interpretation.
He claims (1993: 341) that a sentence like
(106)
America has found a role
stems from a re-analysis of (106)a, where has indicates possession and its object, a
role, is modified by the adjectival PtPPL - the role is in a “state of found-ness” -, to
(106) b, where a role is the object of the verbal syntagm “has found”:
(106) a.
b.
America [VP [V has ] [[NP a role ] [AP found ]]]
have = main
America [VP [ has
have = aux
[ V found ]] [NP a role ]]
4.4 - The status of Friulian with respect to the other
Romance Languages
Going back to Friulian and SI, what examples (103) and (104) suggest is that
the option that was available in Latin for the auxiliary and the PtPPL to have
separate subjects is still available in Friulian. Crucially, the interpretation will rely
on the relative position of vè and the object: the object must immediately follow vè.
This adjacency can also be interpreted as a necessity for vè to be able to assign Acc
92
Case to the object 47.
If this is the case, then we could interpret the fact that this option is available
in Friulian but not in SI 48 as an indication of the “conservativity” of Friulian in
comparison with SI. In other words, Friulian has not yet undergone the change
that SI already has.
This conclusion provides some more support for the idea put forward in the
previous chapter that the PtPPL in Friulian and the PtPPL in SI are specified for
different features, and they derive this difference from a diachronical development.
Friulian and Old Italian represent an earlier stage of the transition which the
PtPPL has undergone. The fact that Friulian allows the present perfect to be
formed with all existing PtPPLs, and not just like early Latin which allowed only
those PtPPLs that could function as adjectives, shows that the same process of
change as the one that has originated the Romance present perfect forms has been
active in Friulian. Given the differences we have noticed with SI, the process has
not completed its cycle in Friulian - or at least in the “Old Friulian” - which
remains “behind” SI as far as the transformation of the PtPPL is concerned.
As we have seen, the structure Aux - PtPPL - DO has the same meaning in
Friulian and SI. This shows that in Friulian, too, the structure has undergone the
semantic change that affected Latin periphrasis like (87) and (88). What Friulian
has not undergone yet, is a full syntactic change, which would result in Friulian not
allowing PtPPL agr with a full pvDP, like SI. This uncompleted change is reflected
in the feature specification of the two PtPPLs.
In Latin, the PtPPL behaved syntactically like an adjective and accordingly it
was specified as a [+N,+V] category. In its development towards the modern
PtPPL, it has ended up as being a [+V] category in SI. The lack of [-N] feature,
typical of all the other verbal categories, formally expresses the potential
In Latin this adjacency was not required given that Latin had a relatively free
word order - due to its rich Case inflection - and tended to place the finite verb at
the end of the sentence. Friulian on the contrary, just like SI, has a fairly strict
word order, and the position of the object affects the meaning of the construction
Aux - PtPPL -DO, as it has been seen.
47
It would be interesting to see whether the corresponding example of (105)
was allowed in Old Italian or Latin, at all.
48
93
“adjectival” nature of the PtPPL in SI - i.e. its ability to agree with a DP which has
been cliticised.
In Friulian, on the other hand, its specification is still [+N,+V], which causes
the PtPPL to be subject to the principle outlined in (81) chapter 3 section 3.3,
showing agreement in [number, gender] with the pvDP. The fact that younger
speakers prefer the “non-agreeing” form of the PtPPL shows that there is a change
in progress: the PtPPL is shifting from a [+N,+V] specification to a [+V] one in
Friulian, and with the old generations gradually dying, the only possibility which
will be available to Friulian speakers in a few decades will be the same as SI: no
PtPPL agr with a full pvDP.
Languages like Rumanian, Portuguese and Spanish have reached the end of
this transition, not allowing for the agreement to be even triggered by cliticisation
of the pvDP. In these languages the PtPPL has probably come to its specification as
a [+V, -N] category, just like other verbal categories.
4.5 -
Conclusion
This chapter has provided some support for the claim made in (71) b. By
investigating the semantic and syntactic structure that the periphrasis had in
Latin, a link has been suggested between the differences noticed between Friulian
and SI and a diachronic development of the PtPPL. While it is a [+V, ~N] category
in SI, where the [N] feature has no value, it remains a [+V, +N] category in
Friulian, just like it was in Latin. Some more evidence for this claim comes from
the possibility - that existed in Latin and which is still available to Friulian but not
to SI - for the subjects of the auxiliary and the subject of the PtPPL not to be coreferential.
Both verb forms, the PtPPL and to have have undergone a change. The
process has involved all Romance languages: while Rumanian, Spanish and
Portuguese have fully accomplished it, SI and French are half way through the
process and Friulian has just started its journey towards the complete
“verbalisation” of the PtPPL.
94
Conclusions
The aim of the present work was two fold. We set off investigating the nature
of PtPPL agr and trying to find an answer to the question “Why are AgrS and AgrO
specified for different features?”. The answer to this question would also, in an
indirect way, throw some light on the tension between agreement as functional
projection and agreement as a relation.
The question was raised by the investigation of Friulian carried out in Paoli
(1996), and it is on this language that the research has been based.
The search for an answer to the question led us first to investigate the
semantic nature of PtPPLs and adjectives: the agreement instantiated by both
categories is specified for [gender, number] features, and it was felt that this could
be due to their semantic affinity. An account of PtPPL agr based on a revised
version of Centineo’s (1996) semantic analysis of PtPPL agr in SI, however, has not
proved powerful enough to explain Friulian data: even with revisions derived from
Tenny’s (1992) and Krifka’s (1989, 1992) theories of aspectuality, a semantic
account could still not account for activity verbs with an indefinite object in
Friulian. Nevertheless, it was made clear that the PtPPL only agrees with its
internal argument, and the semantic affinity between PtPPLs and adjectives was
maintained.
This affinity on a semantic level was interpreted on a syntactic level as
suggesting that the same mechanism responsible for adjectival agreement could be
responsible for PtPPL agr. The mechanism invoked is feature copying, which was
exploited already in the Sixties by Postal (1964). The definition of the agreement
process’s operational domain has proved very significant in capturing differences
in behaviour between PtPPLs and adjectives when agreeing with a list of nominals.
Finally, some speculations were made on the different nature of agreement
specified for [person] and the one specified for [gender] features. Two possible
hypotheses have been suggested: either the appearance of [person] features reflects
the presence of an Agr functional node, or there is no such functional node and
subject-verb agreement is triggered by a [Spec, head] configuration between the
two. The functional projection hosting this configuration would be Tense.
Although these last speculations need further cross-linguistic investigation 95
a suggestion for further research -, they have shed some light on the debate about
the nature of agreement in general. Although linguists feel comfortable with Tense
and Negation, two functional projections that mark clear semantic properties, an
AGRP cannot be accepted as easily. From a functional projection point of view,
AGRPs are semantically empty, agreement being generally thought of as a relation
between two elements, not as a property possessed by an element in isolation.
Using AGRPs has turned the relational character of agreement into a property that
an item acquires by virtue of occupying a particular head position.
Linguists have been aware of this tension, but the matter has been
commented upon mainly in footnotes, not directly addressed. This work has tried
to relate its value with respect to a contained phenomenon, agreement in [gender]
features, to this more theoretical question and to provide some evidence against
the existence of an AgrO projection, possibly even an AgrS one. This latter fact
must receive more attention, and I leave it here open for further research.
96
References
Abney, S P (1987), “The English Noun Phrase in its sentential aspect”, PhD
Dissertation, MIT
♣ Agard, F B (1977), “The Auxiliary in Romance”, in Hagiwara, M P Studies in
Romance Linguistics, Newbury, Rowley, MA
♣ Baker, M C (1985), “The Mirror Principle and Morphosyntactic Explanation”,
Linguistic Inquiry 16:3, 373-415
♣ Baker, M C (1988), Incorporation: A Theory of Grammatical Function Changing,
The Chicago University Press, Chicago
♣ Belletti, A (1981), “Frasi ridotte assolute”, Rivista di Garmmatica Generativa, 6,
3-33
♣ Belletti, A (1990), Generalized Verb Movement: Aspects of Syntax, Rosenberg &
Sellier, Torino
♣ Belletti, A (1994), “Verb positions: evidence from Italian”, in Lightfoot, D &
Hornstein, N (eds), Verb Movement, CUP, Cambridge, 19-40
♣ Benincà, P and Vanelli, L (1984), “Italiano, Veneto, Friuliano: fenomeni sintattici
a confronto”, Rivista Italiana di Dialettologia, Scuola società territorio, VII, numero
unico, 165-194
♣ Benveniste, E (1966), Problèmes de Linguistique Générale, Gallinard, Paris
♣ Bernstein, J (1991), “DPs in French and Walloon: Evidence for Parametric
Variation in Nominal Head Movement”, Probus 3:2, 101-126
♣ Bianchi, V & Figueiredo-Silva, M C (1994), “On Some Properties of
Agreement-Object in Italian and Brazilian Portuguese”, in Mazzola, M L (ed),
Issues and Theory in Romance Linguistics, Georgetown UP, Washington, DC, 181197
♣ Bisetto, A (1994), “Sugli aggettivi in -(x)to”, in Borgato, G (ed), Teoria del
Linguaggio e analisi linguistica, XX incontro di Grammatica Generativa, University
Press Padova, Padova, 63-81
♣ Bouchard, D (1984), On the Content of Empty Categories, Foris Publications,
Dordrecht
♣ Bouchard, D (1985), “A few remarks on past participle agreement in French and
♣
97
Italian”, Ms. University of Texas, in Brown (1988)
♣ Brown, B (1988), “Problems with Past Participle Agreement in French and
Italian Dialects”, in Birdsong, D & Montreuil, J-P (eds), Advances in Romance
Linguistics, Foris Publications, Dordrecht, 51-66
♣ Burzio, L (1981), Intransitive Verbs and Italian Auxiliaries, PhD Thesis, MIT
Cambridge, MA
♣ Burzio, L (1986), Italian Syntax: A Government - Binding Approach, Reidel,
Dordrecht
♣ Centineo, G (1996), “ A lexical Theory of Auxiliary Selection in Italian”, Probus
8.3, 223-271
♣ Chomsky, N (1970), “Remarks on Nominalisation”, in Jacobs, R A & Rosenbaum,
P S (eds), Readings in English Transformational Grammar, Ginn & Co, Waltham,
Mass, 184 - 221
♣ Chomsky, N (1981), Lectures on Government and Binding, Foris Publications,
Dordrecht
♣ Chomsky, N (1986), Barriers, Linguistic Inquiry Monograph 13, The MIT Press,
Cambridge, Mass.
♣ Chomsky, N (1988), “Some notes on economy of derivation and representation,
MIT Working Papers in Linguistics 10, 43-75
♣ Chomsky, N (1991), “Some Notes of Economy of Derivation and Representation”,
in Freidin, R (ed), Principles and Parameters in Comparative Grammar, The MIT
Press, Cambridge, MA, 417-454
♣ Chomsky, N (1995), Minimalist Program, Current Studies in Linguistics, The
MIT Press, Cambridge, MA
♣ Cinque, G (1994), “Sull’ordine relativo di alcune classi di avverbi in italiano e in
francese”, in Borgato, G (ed), Teoria del Linguaggio e analisi linguistica, XX
incontro di Grammatica Generativa, University Press Padova, Padova, 163-177
♣ Cinque, G (1994a), “On the Evidence for Partial N-Movement in the Romance
DP”, in Cinque, G & Koster, J & Pollock, J-Y & Rizzi, L & Zanuttini, R (eds), Paths
Towards Universal Grammar, Georgetown University press, Washington, DC, 85109
♣ Cinque, G (forthcoming), “Adverbs and Functional Heads: A cross-linguistic
98
Perspective”, OUP, Oxford
♣ Cocchi, G (1994), “Sintassi della selezione dell’ausiliare”, Doctorate thesis,
University of Florence
♣ Cocchi, G (1994a), “An explanation for the split in the choice of perfect
auxiliaries”, Probus 6.2/3, 87-102
♣ Cocchi, G (1994b), “Il problema della selezione dell’ausiliare nei tempi
composti”, in Borgato, G (ed), Teoria del Linguaggio e analisi linguistica, XX
incontro di Grammatica Generativa, University Press Padova, Padova, 179-194
♣ Cortés, C (1993), “Catalan participle agreement, auxiliary selection and the
Government Transparency Corollary”, Probus 5.3, 193-240
♣ Denison, D (1993), English Historical Syntax, Longman Linguistics Library,
London & New York
♣ Dixon, R M W (1979), “Ergativity”, Language 55: 59-138 in Foley & van Valin
♣ Dowty, D R (1979), Word Meaning and Montague Grammar : The Semantics of
Verbs and Times in Generative Semantics and in Montagues’s PTQ, Reidel,
Dordrecht, Holland
♣Emonds, J (1978), “The verbal Complex V’-V in French”, Linguistic Inquiry, 9
151-175
♣ Fassi Fehri, A (1989), “Generalised IP structure, Case and VS Word Order”, in
Laka, I & Mahajan, A (eds), MIT Working Papers in Linguistics 10, 75-113
♣ Fillmore, C (1968), “The Case for Case”, in Bach E. & Harms, R (eds), Universals
in Linguistic Theory, pp 1-188, Holtz, Rinehart & Winston, New York
♣ Fodor, J A (1970), “Three Reasons for Not Deriving 'Kill' from 'Cause to Die'”,
Linguistic Inquiry, 1, 429-38
♣ Foley, W A, & van Valin, R D (1984), Functional Syntax and Universal Grammar,
CUP, Cambridge
♣ Giorgi, A & Longobardi, G (1990), The syntax of noun phrases: Configuration,
parameters and empty categories, CUP, Cambridge
♣ Giorgi, A & Pianesi, F (1991), “Toward a Syntax of Temporal Representations”,
Probus 3.2, 187-213
♣ Grimshaw, J (1991), “Extended Projection”, Unpublished Manuscript, Brandeis
University, Waltham, MA
99
♣ Gruber, J (1965), Studies in lexical relations, MIT, Dissertation, in Foley & van
Valin
♣ Guéron, J & Hoekstra, T (1988), “T-Chains and the Constituent Structure of
Auxiliaries”, in Cardinaletti, A & Cinque, G & Giusti, G (eds), Constituent
Structure, Foris Publications, Dordrecht, 35-99
♣ Heath, J (1975), “Some functional relationship in grammar”, Language 51: 89104 in Foley & van Valin
♣ Hjelmslev, L (1943), "Prolegomena to a Theory of Language" , Indiana University
Press, Bloomington
♣ Holmberg, A and Platzack, C (1995), The Role of Inflection in Scandinavian
Syntax, OUP, Oxford
♣ Hyams, N (1981), “The Choice of Auxiliary and Agreement in Italian”, in Burke,
V & Pustejovsky, J (eds), Proceedings of the Eleventh Annual Meeting of the North
Eastern Linguistic Society, Amherst, MA, 143-154
♣ Iatridou, S (1990), “About AGR(P)”, Linguistic Inquiry 24, 551-577
♣ Jakobson, R (1936), "Beitrag zur allgemeinen Kasuslehre", Travaux de Cercle
Linguistique de Prague 6, 240-288 in Dowty (1979)
♣ Kayne, R (1975), French Syntax, The MIT Press, Cambridge, Mass.
♣ Kayne, R (1981), “ECP extensions”, Linguistic Inquiry 12, 93-133, reprinted in
Kayne 1984.
♣ Kayne, R (1984), Connectedness and Binary Branching, Foris Publications,
Dordrecht
♣ Kayne, R (1985), “L’acord du participe passè en français et en italien”, Modèles
Linguistiques, VII, I, 73-91
♣ Kayne, R (1989), “Null Subjects and Clitic Climbing”, in Jaeggli, O & Safir, K J
(eds), The Null Subject Parameter, Foris Publications, Dordrecht, 239-261
♣ Kayne, R (1991), “Romance Clitics, Verb Movement and PRO”, Linguistic Inquiry
22, 647-686
♣ Kayne, R (1993), “Toward a Modular Theory of Auxiliary Selection”, Studia
Linguistica 47:1, 3-31
♣ Kayne, R (1994), “The Antisymmetry of Syntax”, The MIT Press, Cambridge,
Mass.
100
♣ Klima, E (1964), “Negation in English”, in Fodor, J & Katz, J (eds), The Structure
of Language: Readings in the Philosophy of Language, Englewood Cliffs, Prentice
Hall, NJ
♣ Koopman, H & Sportiche, D (1988), “Subjects”, Unpublished Manuscript,
University of California, LA
♣ Krifka, M (1989), “Nominal Reference, Temporal Constitution and
Quantification in Event Semantics”, in Bartsch, R & van Benthem, J & van Emde
Boas, P (eds), Semantics and Contextual Expression, Foris Publications, Dordrecht
♣ Krifka, M (1992), “Thematic Relations as Links between Nominal Reference and
Temporal Constitution”, in Sag, I A & Szabolcsi, A (eds), Lexical Matters, CSLI,
USA, 29-54
♣ Lamarche, J (1991), “Problems for N° movement to NumP”, Probus 3:2, 215-236
♣Lakoff, G (1965), "On the nature of Syntactic Regularity", Doctoral dissertation,
Indiana University
♣Lefebvre, C (1988), “Past Participle Agreement in French: Agreement = Case” in
Birdsong, D & Montreuil, J-P (eds), Advances in Romance Linguistics, Foris
Publications, Dordrecht, 233-251
♣ Lehmann, C (1982), “Universal and Typological aspects of agreement”, in Seiler,
H & Stachowiak, F J (eds), Apprehension: das sprachliche Erfassen von
Gegenständen, Teil II, Gunter Narr,Tübingen, 201-267
♣Lehmann, C (1988), “On the Function of Agreement”, in Barlow, M & Ferguson, C
A Agreement in Natural Language : Approaches, Theories, Descriptions, CSLI,
United States, 55-65
♣ Leone, A (1954), “A proposito degli ausiliari”, Lingua Nostra 15: 127-131
♣ Leone, A (1970), “Una regola per gli ausiliari”, Lingua Nostra 31: 24-30
♣ Lepschy, A L & Lepschy, G (1981), La Lingua Italiana, Bompiani, Milano
♣ Lois, X (1990), “Auxiliary Selection and Past Participle Agreement in Romance”,
Probus 2.2, 233-255
♣ McCawley, J D (1968), "The Role of Semantics in a grammar", in Bach, E &
Harms, R (eds), Universal in Linguistic Theory, Holt, Rinehart & Winston, New
York, 124-169
♣ Paoli, S (1996), The “Agreement Parameter”: Evidence from a Northern Italian
101
Dialect, Undergraduate Dissertation, Manuscript, University of York, York
♣ Parisi, D (1976), “The Past Participle”, Italian Linguistics 1, 77-106
♣ Perlmutter, D M (1978), “Impersonal Passives and the unaccusative hypothesis”,
Berkeley Linguistic Society, 4:157-189
♣ Perlmutter, D M (1980), “Relational Grammar”, in Moravcsik, E A & Wirth, J R
(eds), Current Approaches to Syntax, Number 13: 195-229, Academic Press, New
York
♣ Picallo, C (1991), “Nominals and Nominalisation in Catalan”, Probus 3, 279-316
♣ Pinkster, H (1987), “The Strategy and Chronology of the Development of Future
and Perfect Tense Auxiliaries in Latin”, in Harris, M & Ramat, P (eds), Historical
Development of Auxiliaries, Mouton de Gruyter, Berlin, 193-223
♣ Poletto, C (1992), “La sintassi del soggetto nei dialetti italiani settentrionali”,
PhD Thesis, University of Venice
♣ Postal, P (1964), Constituent Structure: A Study of Contemporary Models of
Syntactic Description, Indiana University, Bloomington, The Netherlands
♣ Pollock, J-Y (1989), “Verb Movement, Universal Grammar and the Structure of
IP”, Linguistic Inquiry 20.3, 365-424
♣ Quine, W V (1960), Word and Object, The MIT Press, Cambridge, MA
♣ Ramat, P (1982), “Ein Beispel von “reanalysis”, typoligisch betrachtet”, FoL 16,
365-383, in Pinkster, H
♣ Randall, J G (1986), Learning Latin, Francis Cairns Publications, University of
Leeds
♣ Reichenbach, H (1966), Elements of Symbolic Logics, MacMillian, New York, In
Giorgi & Pianesi (1991)
♣ Ritter, E (1993), “Where’s Gender?”, Linguistic Inquiry, 24:4, 795-803
♣ Ritter E (1991), “Two Functional Categories in NPs: Evidence from Modern
Hebrew”, in Rothstein, S D (ed), Perspectives on Phrase Structure: Heads and
Licensing, Syntax and Semantics 25, 37-62
♣ Rizzi, L (1982), Italian Syntax, Foris Publications, Dordrecht
♣ Rizzi, L (1986), “On the status of subject clitics in Romance”, in Jaeggli, O &
Silva-Corvalán, C (eds), Studies in Romance Linguistics, Foris Publications,
Dordrecht, 391-419
102
Rohlfs, G (1969), Grammatica Storica della Lingua Italiana, Giulio Einaudi
Editore, Torino
♣ Salvi, G (1987), “Syntactic Restructuring in the Evolution of Romance
Auxiliaries” in Harris, M & Ramat, P (eds), Historical Development of Auxiliaries,
Mouton de Gruyter, Berlin, 225-236
♣ Sportiche, D(1996), “Clitic Constructions”, in Rooryck, J & Zaring, L (eds),
Phrase Structure and the Lexicon, Kluwer, Dordrecht, 213-276
♣ Stowell, T (1982), “Subject across categories”, The Linguistic Review 2, 285-312
♣
Szabolcsi, A (1981), “The possessive Construction in Hungarian: A
Configurational Category in a Non-Configurational Language”, Acta Linguistica
Academia Scientiarum Hungaricae 31:1-4, 261-289
♣ Szabolcsi, A (1987), “Functional Categories in the NP”, in Renesei (eds),
Approaches to Hungarian 2, 167-189
♣ Szabolcsi, A (1994), “The Noun Phrase”, in Kiefer, F & Kiss, K É (eds), Syntax
and Semantics 27: The Syntactic Structure of Hungarian, Academic Press, 179-274
♣ Taraldsen, T (1987), “Clitic/Particle agreement and Auxiliary Alternation in
Romance”, In Neidle, C & Nuñez Cedeño, R A (eds), Studies in Romance
Languages, Foris Publications, Dordrecht, 263-281
♣ Tenny, C (1992), “The Aspectual Interface Hypothesis”, in Sag, I A & Szabolcsi, A
(eds), Lexical Matters, CSLI, USA, 1-28
♣ van Voorst, J (1988), Event Structure, Current Issues in Linguistic Theory 59,
John Benjamins publishing Company, Amsterdam
♣ Vanelli, L (1997), “Friuli”, in Maiden, M & Parry, M (eds), The dialects of Italy,
Routledge, London, 279-285
♣ Vendler, Z (1967), Linguistics and Philosophy, Cornell University Press, Ithaca
♣ Vikner, S & Sprouse, R A (1988), “Parameters of Have/Be Selection in Germanic
and Romance”, NELS 18/2: 523-537
♣ Vikner, S (1991), “Be Is Selected over Have If and Only If It Is Part of an
A-chain”, in Abraham, W & Kosmeijer, W & Reuland, E (eds), Issues in Germanic
Syntax, Mouton de Gruyter, Berlin, 365-381
♣ Vincent, N (1982), “The development of the auxiliaries habere and esse in
Romance”, in Vincent, N & Harris, M (eds), Studies in the Romance Verb, Croom
♣
103
Helm, London, 71-96
♣ Williams, E (1981), “Argument Structure and Morphology”, Linguistics Review,
1, 82-114
♣ Williams, E (1994), “A reinterpretation of evidence for verb movement in
French”, in Lightfoot, D & Hornstein, N (eds), Verb Movement, CUP, Cambridge,
189-205
♣ Zanuttini, R (1991), “Syntactic properties of sentential negation: a comparative
study of Romance languages”, PhD Thesis, University of Pennsylvania
104